![Page 1: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/1.jpg)
V. Approx. Algorithms: Travelling Salesman ProblemThomas Sauerwald
Easter 2020
![Page 2: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/2.jpg)
Outline
Introduction
General TSP
Metric TSP
V. Travelling Salesman Problem Introduction 2
![Page 3: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/3.jpg)
33 city contest (1964)
Traveling Salesman 12
rather simple methods could be found to yield a tour much nearer optimal than 30
percent. However, even a few percent gain would be well worth-while in some
cases, so the problem does seem to have practical importance as well as
mathematical interest. (p. 65)
Thus Flood realized that the Nearest Neighbor method is not a good estimate of the TSP
but it created a decent first solution.
In 1962 a contest brought the TSP national recognition through a contest given by
Proctor and Gamble. A flyer of the contest is pictured below.
The traveling salesman problem recently achieved national prominence when a
soap company used it as the basis of a promotional contest. Prizes up to $10,000
V. Travelling Salesman Problem Introduction 3
![Page 4: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/4.jpg)
532 cities (1987 [Padberg, Rinaldi])
V. Travelling Salesman Problem Introduction 4
![Page 5: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/5.jpg)
13,509 cities (1999 [Applegate, Bixby, Chavatal, Cook])
V. Travelling Salesman Problem Introduction 5
![Page 6: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/6.jpg)
The Traveling Salesman Problem (TSP)
Given a set of cities along with the cost of travel between them, find thecheapest route visiting all cities and returning to your starting point.
Given: A complete undirected graph G = (V ,E) withnonnegative integer cost c(u, v) for each edge (u, v) ∈ E
Goal: Find a hamiltonian cycle of G with minimum cost.
Formal Definition
Solution space consists of at most n! possible tours!
Actually the right number is (n − 1)!/2
3
1
2 1
4
3
3 + 2 + 1 + 3 = 92 + 4 + 1 + 1 = 8
Metric TSP: costs satisfy triangle inequality:
∀u, v ,w ∈ V : c(u,w) ≤ c(u, v) + c(v ,w).
Euclidean TSP: cities are points in the Euclidean space, costs areequal to their (rounded) Euclidean distance
Special InstancesEven this version is
NP hard (Ex. 35.2-2)
V. Travelling Salesman Problem Introduction 6
![Page 7: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/7.jpg)
The Traveling Salesman Problem (TSP)
Given a set of cities along with the cost of travel between them, find thecheapest route visiting all cities and returning to your starting point.
Given: A complete undirected graph G = (V ,E) withnonnegative integer cost c(u, v) for each edge (u, v) ∈ E
Goal: Find a hamiltonian cycle of G with minimum cost.
Formal Definition
Solution space consists of at most n! possible tours!
Actually the right number is (n − 1)!/2
3
1
2 1
4
3
3 + 2 + 1 + 3 = 92 + 4 + 1 + 1 = 8
Metric TSP: costs satisfy triangle inequality:
∀u, v ,w ∈ V : c(u,w) ≤ c(u, v) + c(v ,w).
Euclidean TSP: cities are points in the Euclidean space, costs areequal to their (rounded) Euclidean distance
Special InstancesEven this version is
NP hard (Ex. 35.2-2)
V. Travelling Salesman Problem Introduction 6
![Page 8: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/8.jpg)
The Traveling Salesman Problem (TSP)
Given a set of cities along with the cost of travel between them, find thecheapest route visiting all cities and returning to your starting point.
Given: A complete undirected graph G = (V ,E) withnonnegative integer cost c(u, v) for each edge (u, v) ∈ E
Goal: Find a hamiltonian cycle of G with minimum cost.
Formal Definition
Solution space consists of at most n! possible tours!
Actually the right number is (n − 1)!/2
3
1
2 1
4
3
3 + 2 + 1 + 3 = 92 + 4 + 1 + 1 = 8
Metric TSP: costs satisfy triangle inequality:
∀u, v ,w ∈ V : c(u,w) ≤ c(u, v) + c(v ,w).
Euclidean TSP: cities are points in the Euclidean space, costs areequal to their (rounded) Euclidean distance
Special InstancesEven this version is
NP hard (Ex. 35.2-2)
V. Travelling Salesman Problem Introduction 6
![Page 9: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/9.jpg)
The Traveling Salesman Problem (TSP)
Given a set of cities along with the cost of travel between them, find thecheapest route visiting all cities and returning to your starting point.
Given: A complete undirected graph G = (V ,E) withnonnegative integer cost c(u, v) for each edge (u, v) ∈ E
Goal: Find a hamiltonian cycle of G with minimum cost.
Formal Definition
Solution space consists of at most n! possible tours!
Actually the right number is (n − 1)!/2
3
1
2 1
4
3
3 + 2 + 1 + 3 = 92 + 4 + 1 + 1 = 8
Metric TSP: costs satisfy triangle inequality:
∀u, v ,w ∈ V : c(u,w) ≤ c(u, v) + c(v ,w).
Euclidean TSP: cities are points in the Euclidean space, costs areequal to their (rounded) Euclidean distance
Special InstancesEven this version is
NP hard (Ex. 35.2-2)
V. Travelling Salesman Problem Introduction 6
![Page 10: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/10.jpg)
The Traveling Salesman Problem (TSP)
Given a set of cities along with the cost of travel between them, find thecheapest route visiting all cities and returning to your starting point.
Given: A complete undirected graph G = (V ,E) withnonnegative integer cost c(u, v) for each edge (u, v) ∈ E
Goal: Find a hamiltonian cycle of G with minimum cost.
Formal Definition
Solution space consists of at most n! possible tours!
Actually the right number is (n − 1)!/2
3
1
2 1
4
3
3 + 2 + 1 + 3 = 92 + 4 + 1 + 1 = 8
Metric TSP: costs satisfy triangle inequality:
∀u, v ,w ∈ V : c(u,w) ≤ c(u, v) + c(v ,w).
Euclidean TSP: cities are points in the Euclidean space, costs areequal to their (rounded) Euclidean distance
Special InstancesEven this version is
NP hard (Ex. 35.2-2)
V. Travelling Salesman Problem Introduction 6
![Page 11: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/11.jpg)
The Traveling Salesman Problem (TSP)
Given a set of cities along with the cost of travel between them, find thecheapest route visiting all cities and returning to your starting point.
Given: A complete undirected graph G = (V ,E) withnonnegative integer cost c(u, v) for each edge (u, v) ∈ E
Goal: Find a hamiltonian cycle of G with minimum cost.
Formal Definition
Solution space consists of at most n! possible tours!
Actually the right number is (n − 1)!/2
3
1
2 1
4
3
3 + 2 + 1 + 3 = 9
2 + 4 + 1 + 1 = 8
Metric TSP: costs satisfy triangle inequality:
∀u, v ,w ∈ V : c(u,w) ≤ c(u, v) + c(v ,w).
Euclidean TSP: cities are points in the Euclidean space, costs areequal to their (rounded) Euclidean distance
Special InstancesEven this version is
NP hard (Ex. 35.2-2)
V. Travelling Salesman Problem Introduction 6
![Page 12: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/12.jpg)
The Traveling Salesman Problem (TSP)
Given a set of cities along with the cost of travel between them, find thecheapest route visiting all cities and returning to your starting point.
Given: A complete undirected graph G = (V ,E) withnonnegative integer cost c(u, v) for each edge (u, v) ∈ E
Goal: Find a hamiltonian cycle of G with minimum cost.
Formal Definition
Solution space consists of at most n! possible tours!
Actually the right number is (n − 1)!/2
3
1
2 1
4
3
3 + 2 + 1 + 3 = 9
2 + 4 + 1 + 1 = 8
Metric TSP: costs satisfy triangle inequality:
∀u, v ,w ∈ V : c(u,w) ≤ c(u, v) + c(v ,w).
Euclidean TSP: cities are points in the Euclidean space, costs areequal to their (rounded) Euclidean distance
Special InstancesEven this version is
NP hard (Ex. 35.2-2)
V. Travelling Salesman Problem Introduction 6
![Page 13: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/13.jpg)
The Traveling Salesman Problem (TSP)
Given a set of cities along with the cost of travel between them, find thecheapest route visiting all cities and returning to your starting point.
Given: A complete undirected graph G = (V ,E) withnonnegative integer cost c(u, v) for each edge (u, v) ∈ E
Goal: Find a hamiltonian cycle of G with minimum cost.
Formal Definition
Solution space consists of at most n! possible tours!
Actually the right number is (n − 1)!/2
3
1
2 1
4
3
3 + 2 + 1 + 3 = 9
2 + 4 + 1 + 1 = 8
Metric TSP: costs satisfy triangle inequality:
∀u, v ,w ∈ V : c(u,w) ≤ c(u, v) + c(v ,w).
Euclidean TSP: cities are points in the Euclidean space, costs areequal to their (rounded) Euclidean distance
Special InstancesEven this version is
NP hard (Ex. 35.2-2)
V. Travelling Salesman Problem Introduction 6
![Page 14: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/14.jpg)
The Traveling Salesman Problem (TSP)
Given a set of cities along with the cost of travel between them, find thecheapest route visiting all cities and returning to your starting point.
Given: A complete undirected graph G = (V ,E) withnonnegative integer cost c(u, v) for each edge (u, v) ∈ E
Goal: Find a hamiltonian cycle of G with minimum cost.
Formal Definition
Solution space consists of at most n! possible tours!
Actually the right number is (n − 1)!/2
3
1
2 1
4
3
3 + 2 + 1 + 3 = 9
2 + 4 + 1 + 1 = 8
Metric TSP: costs satisfy triangle inequality:
∀u, v ,w ∈ V : c(u,w) ≤ c(u, v) + c(v ,w).
Euclidean TSP: cities are points in the Euclidean space, costs areequal to their (rounded) Euclidean distance
Special InstancesEven this version is
NP hard (Ex. 35.2-2)
V. Travelling Salesman Problem Introduction 6
![Page 15: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/15.jpg)
The Traveling Salesman Problem (TSP)
Given a set of cities along with the cost of travel between them, find thecheapest route visiting all cities and returning to your starting point.
Given: A complete undirected graph G = (V ,E) withnonnegative integer cost c(u, v) for each edge (u, v) ∈ E
Goal: Find a hamiltonian cycle of G with minimum cost.
Formal Definition
Solution space consists of at most n! possible tours!
Actually the right number is (n − 1)!/2
3
1
2 1
4
3
3 + 2 + 1 + 3 = 9
2 + 4 + 1 + 1 = 8
Metric TSP: costs satisfy triangle inequality:
∀u, v ,w ∈ V : c(u,w) ≤ c(u, v) + c(v ,w).
Euclidean TSP: cities are points in the Euclidean space, costs areequal to their (rounded) Euclidean distance
Special Instances
Even this version isNP hard (Ex. 35.2-2)
V. Travelling Salesman Problem Introduction 6
![Page 16: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/16.jpg)
The Traveling Salesman Problem (TSP)
Given a set of cities along with the cost of travel between them, find thecheapest route visiting all cities and returning to your starting point.
Given: A complete undirected graph G = (V ,E) withnonnegative integer cost c(u, v) for each edge (u, v) ∈ E
Goal: Find a hamiltonian cycle of G with minimum cost.
Formal Definition
Solution space consists of at most n! possible tours!
Actually the right number is (n − 1)!/2
3
1
2 1
4
3
3 + 2 + 1 + 3 = 9
2 + 4 + 1 + 1 = 8
Metric TSP: costs satisfy triangle inequality:
∀u, v ,w ∈ V : c(u,w) ≤ c(u, v) + c(v ,w).
Euclidean TSP: cities are points in the Euclidean space, costs areequal to their (rounded) Euclidean distance
Special Instances
Even this version isNP hard (Ex. 35.2-2)
V. Travelling Salesman Problem Introduction 6
![Page 17: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/17.jpg)
The Traveling Salesman Problem (TSP)
Given a set of cities along with the cost of travel between them, find thecheapest route visiting all cities and returning to your starting point.
Given: A complete undirected graph G = (V ,E) withnonnegative integer cost c(u, v) for each edge (u, v) ∈ E
Goal: Find a hamiltonian cycle of G with minimum cost.
Formal Definition
Solution space consists of at most n! possible tours!
Actually the right number is (n − 1)!/2
3
1
2 1
4
3
3 + 2 + 1 + 3 = 9
2 + 4 + 1 + 1 = 8
Metric TSP: costs satisfy triangle inequality:
∀u, v ,w ∈ V : c(u,w) ≤ c(u, v) + c(v ,w).
Euclidean TSP: cities are points in the Euclidean space, costs areequal to their (rounded) Euclidean distance
Special InstancesEven this version is
NP hard (Ex. 35.2-2)
V. Travelling Salesman Problem Introduction 6
![Page 18: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/18.jpg)
The Traveling Salesman Problem (TSP)
Given a set of cities along with the cost of travel between them, find thecheapest route visiting all cities and returning to your starting point.
Given: A complete undirected graph G = (V ,E) withnonnegative integer cost c(u, v) for each edge (u, v) ∈ E
Goal: Find a hamiltonian cycle of G with minimum cost.
Formal Definition
Solution space consists of at most n! possible tours!
Actually the right number is (n − 1)!/2
3
1
2 1
4
3
3 + 2 + 1 + 3 = 9
2 + 4 + 1 + 1 = 8
Metric TSP: costs satisfy triangle inequality:
∀u, v ,w ∈ V : c(u,w) ≤ c(u, v) + c(v ,w).
Euclidean TSP: cities are points in the Euclidean space, costs areequal to their (rounded) Euclidean distance
Special InstancesEven this version is
NP hard (Ex. 35.2-2)
V. Travelling Salesman Problem Introduction 6
![Page 19: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/19.jpg)
History of the TSP problem (1954)
Dantzig, Fulkerson and Johnson found an optimal tour through 42 cities.
http://www.math.uwaterloo.ca/tsp/history/img/dantzig_big.html
V. Travelling Salesman Problem Introduction 7
![Page 20: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/20.jpg)
The Dantzig-Fulkerson-Johnson Method
1. Create a linear program (variable x(u, v) = 1 iff tour goes between u and v )
2. Solve the linear program. If the solution is integral and forms a tour, stop.Otherwise find a new constraint to add (cutting plane)
0 1 2 3 4 5 6 7 8 9
1
2
3
4
5
max 13 x + y
4x1 + 9x2 ≤ 36
2x1 − 9x2 ≤ −27
x1
x2
x2 ≤ 3
Additional constraint to cutthe solution space of the LP
(1.5, 3.3)(2.25, 3)
(2, 3)
More cuts are needed to find integral solution
V. Travelling Salesman Problem Introduction 8
![Page 21: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/21.jpg)
The Dantzig-Fulkerson-Johnson Method
1. Create a linear program (variable x(u, v) = 1 iff tour goes between u and v )2. Solve the linear program. If the solution is integral and forms a tour, stop.
Otherwise find a new constraint to add (cutting plane)
0 1 2 3 4 5 6 7 8 9
1
2
3
4
5
max 13 x + y
4x1 + 9x2 ≤ 36
2x1 − 9x2 ≤ −27
x1
x2
x2 ≤ 3
Additional constraint to cutthe solution space of the LP
(1.5, 3.3)(2.25, 3)
(2, 3)
More cuts are needed to find integral solution
V. Travelling Salesman Problem Introduction 8
![Page 22: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/22.jpg)
The Dantzig-Fulkerson-Johnson Method
1. Create a linear program (variable x(u, v) = 1 iff tour goes between u and v )2. Solve the linear program. If the solution is integral and forms a tour, stop.
Otherwise find a new constraint to add (cutting plane)
0 1 2 3 4 5 6 7 8 9
1
2
3
4
5
max 13 x + y
4x1 + 9x2 ≤ 36
2x1 − 9x2 ≤ −27
x1
x2
x2 ≤ 3
Additional constraint to cutthe solution space of the LP
(1.5, 3.3)(2.25, 3)
(2, 3)
More cuts are needed to find integral solution
V. Travelling Salesman Problem Introduction 8
![Page 23: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/23.jpg)
The Dantzig-Fulkerson-Johnson Method
1. Create a linear program (variable x(u, v) = 1 iff tour goes between u and v )2. Solve the linear program. If the solution is integral and forms a tour, stop.
Otherwise find a new constraint to add (cutting plane)
0 1 2 3 4 5 6 7 8 9
1
2
3
4
5
max 13 x + y
4x1 + 9x2 ≤ 36
2x1 − 9x2 ≤ −27
x1
x2
x2 ≤ 3
Additional constraint to cutthe solution space of the LP
(1.5, 3.3)(2.25, 3)
(2, 3)
More cuts are needed to find integral solution
V. Travelling Salesman Problem Introduction 8
![Page 24: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/24.jpg)
The Dantzig-Fulkerson-Johnson Method
1. Create a linear program (variable x(u, v) = 1 iff tour goes between u and v )2. Solve the linear program. If the solution is integral and forms a tour, stop.
Otherwise find a new constraint to add (cutting plane)
0 1 2 3 4 5 6 7 8 9
1
2
3
4
5
max 13 x + y
4x1 + 9x2 ≤ 36
2x1 − 9x2 ≤ −27
x1
x2
x2 ≤ 3
Additional constraint to cutthe solution space of the LP
(1.5, 3.3)
(2.25, 3)
(2, 3)
More cuts are needed to find integral solution
V. Travelling Salesman Problem Introduction 8
![Page 25: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/25.jpg)
The Dantzig-Fulkerson-Johnson Method
1. Create a linear program (variable x(u, v) = 1 iff tour goes between u and v )2. Solve the linear program. If the solution is integral and forms a tour, stop.
Otherwise find a new constraint to add (cutting plane)
0 1 2 3 4 5 6 7 8 9
1
2
3
4
5
max 13 x + y
4x1 + 9x2 ≤ 36
2x1 − 9x2 ≤ −27
x1
x2
x2 ≤ 3
Additional constraint to cutthe solution space of the LP
(1.5, 3.3)
(2.25, 3)
(2, 3)
More cuts are needed to find integral solution
V. Travelling Salesman Problem Introduction 8
![Page 26: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/26.jpg)
The Dantzig-Fulkerson-Johnson Method
1. Create a linear program (variable x(u, v) = 1 iff tour goes between u and v )2. Solve the linear program. If the solution is integral and forms a tour, stop.
Otherwise find a new constraint to add (cutting plane)
0 1 2 3 4 5 6 7 8 9
1
2
3
4
5
max 13 x + y
4x1 + 9x2 ≤ 36
2x1 − 9x2 ≤ −27
x1
x2
x2 ≤ 3
Additional constraint to cutthe solution space of the LP
(1.5, 3.3)
(2.25, 3)
(2, 3)
More cuts are needed to find integral solution
V. Travelling Salesman Problem Introduction 8
![Page 27: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/27.jpg)
The Dantzig-Fulkerson-Johnson Method
1. Create a linear program (variable x(u, v) = 1 iff tour goes between u and v )2. Solve the linear program. If the solution is integral and forms a tour, stop.
Otherwise find a new constraint to add (cutting plane)
0 1 2 3 4 5 6 7 8 9
1
2
3
4
5
max 13 x + y
4x1 + 9x2 ≤ 36
2x1 − 9x2 ≤ −27
x1
x2
x2 ≤ 3
Additional constraint to cutthe solution space of the LP
(1.5, 3.3)
(2.25, 3)
(2, 3)
More cuts are needed to find integral solution
V. Travelling Salesman Problem Introduction 8
![Page 28: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/28.jpg)
The Dantzig-Fulkerson-Johnson Method
1. Create a linear program (variable x(u, v) = 1 iff tour goes between u and v )2. Solve the linear program. If the solution is integral and forms a tour, stop.
Otherwise find a new constraint to add (cutting plane)
0 1 2 3 4 5 6 7 8 9
1
2
3
4
5
max 13 x + y
4x1 + 9x2 ≤ 36
2x1 − 9x2 ≤ −27
x1
x2
x2 ≤ 3
Additional constraint to cutthe solution space of the LP
(1.5, 3.3)
(2.25, 3)
(2, 3)
More cuts are needed to find integral solution
V. Travelling Salesman Problem Introduction 8
![Page 29: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/29.jpg)
The Dantzig-Fulkerson-Johnson Method
1. Create a linear program (variable x(u, v) = 1 iff tour goes between u and v )2. Solve the linear program. If the solution is integral and forms a tour, stop.
Otherwise find a new constraint to add (cutting plane)
0 1 2 3 4 5 6 7 8 9
1
2
3
4
5
max 13 x + y
4x1 + 9x2 ≤ 36
2x1 − 9x2 ≤ −27
x1
x2
x2 ≤ 3
Additional constraint to cutthe solution space of the LP
(1.5, 3.3)(2.25, 3)
(2, 3)
More cuts are needed to find integral solution
V. Travelling Salesman Problem Introduction 8
![Page 30: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/30.jpg)
Outline
Introduction
General TSP
Metric TSP
V. Travelling Salesman Problem General TSP 9
![Page 31: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/31.jpg)
Hardness of Approximation
If P 6= NP, then for any constant ρ ≥ 1, there is no polynomial-time ap-proximation algorithm with approximation ratio ρ for the general TSP.
Theorem 35.3
Proof:Idea: Reduction from the hamiltonian-cycle problem.
Let G = (V ,E) be an instance of the hamiltonian-cycle problemLet G′ = (V ,E ′) be a complete graph with costs for each (u, v) ∈ E ′:
c(u, v) =
{1 if (u, v) ∈ E ,ρ|V |+ 1 otherwise.
If G has a hamiltonian cycle H, then (G′, c) contains a tour of cost |V |If G does not have a hamiltonian cycle, then any tour T must use some edge 6∈ E ,
⇒ c(T ) ≥ (ρ|V |+ 1) + (|V | − 1)
= (ρ+ 1)|V |.
Gap of ρ+ 1 between tours which are using only edges in G and those which don’tρ-Approximation of TSP in G′ computes hamiltonian cycle in G (if one exists)
Large weight will renderthis edge useless!
Can create representations of G′ andc in time polynomial in |V | and |E |!
G = (V ,E)
Reduction
G′ = (V ,E ′)
1
1
1
1ρ · 4 + 1
1�1ρ · 4 + 1
V. Travelling Salesman Problem General TSP 10
![Page 32: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/32.jpg)
Hardness of Approximation
If P 6= NP, then for any constant ρ ≥ 1, there is no polynomial-time ap-proximation algorithm with approximation ratio ρ for the general TSP.
Theorem 35.3
Proof:
Idea: Reduction from the hamiltonian-cycle problem.
Let G = (V ,E) be an instance of the hamiltonian-cycle problemLet G′ = (V ,E ′) be a complete graph with costs for each (u, v) ∈ E ′:
c(u, v) =
{1 if (u, v) ∈ E ,ρ|V |+ 1 otherwise.
If G has a hamiltonian cycle H, then (G′, c) contains a tour of cost |V |If G does not have a hamiltonian cycle, then any tour T must use some edge 6∈ E ,
⇒ c(T ) ≥ (ρ|V |+ 1) + (|V | − 1)
= (ρ+ 1)|V |.
Gap of ρ+ 1 between tours which are using only edges in G and those which don’tρ-Approximation of TSP in G′ computes hamiltonian cycle in G (if one exists)
Large weight will renderthis edge useless!
Can create representations of G′ andc in time polynomial in |V | and |E |!
G = (V ,E)
Reduction
G′ = (V ,E ′)
1
1
1
1ρ · 4 + 1
1�1ρ · 4 + 1
V. Travelling Salesman Problem General TSP 10
![Page 33: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/33.jpg)
Hardness of Approximation
If P 6= NP, then for any constant ρ ≥ 1, there is no polynomial-time ap-proximation algorithm with approximation ratio ρ for the general TSP.
Theorem 35.3
Proof:Idea: Reduction from the hamiltonian-cycle problem.
Let G = (V ,E) be an instance of the hamiltonian-cycle problemLet G′ = (V ,E ′) be a complete graph with costs for each (u, v) ∈ E ′:
c(u, v) =
{1 if (u, v) ∈ E ,ρ|V |+ 1 otherwise.
If G has a hamiltonian cycle H, then (G′, c) contains a tour of cost |V |If G does not have a hamiltonian cycle, then any tour T must use some edge 6∈ E ,
⇒ c(T ) ≥ (ρ|V |+ 1) + (|V | − 1)
= (ρ+ 1)|V |.
Gap of ρ+ 1 between tours which are using only edges in G and those which don’tρ-Approximation of TSP in G′ computes hamiltonian cycle in G (if one exists)
Large weight will renderthis edge useless!
Can create representations of G′ andc in time polynomial in |V | and |E |!
G = (V ,E)
Reduction
G′ = (V ,E ′)
1
1
1
1ρ · 4 + 1
1�1ρ · 4 + 1
V. Travelling Salesman Problem General TSP 10
![Page 34: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/34.jpg)
Hardness of Approximation
If P 6= NP, then for any constant ρ ≥ 1, there is no polynomial-time ap-proximation algorithm with approximation ratio ρ for the general TSP.
Theorem 35.3
Proof:Idea: Reduction from the hamiltonian-cycle problem.
Let G = (V ,E) be an instance of the hamiltonian-cycle problem
Let G′ = (V ,E ′) be a complete graph with costs for each (u, v) ∈ E ′:
c(u, v) =
{1 if (u, v) ∈ E ,ρ|V |+ 1 otherwise.
If G has a hamiltonian cycle H, then (G′, c) contains a tour of cost |V |If G does not have a hamiltonian cycle, then any tour T must use some edge 6∈ E ,
⇒ c(T ) ≥ (ρ|V |+ 1) + (|V | − 1)
= (ρ+ 1)|V |.
Gap of ρ+ 1 between tours which are using only edges in G and those which don’tρ-Approximation of TSP in G′ computes hamiltonian cycle in G (if one exists)
Large weight will renderthis edge useless!
Can create representations of G′ andc in time polynomial in |V | and |E |!
G = (V ,E)
Reduction
G′ = (V ,E ′)
1
1
1
1ρ · 4 + 1
1�1ρ · 4 + 1
V. Travelling Salesman Problem General TSP 10
![Page 35: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/35.jpg)
Hardness of Approximation
If P 6= NP, then for any constant ρ ≥ 1, there is no polynomial-time ap-proximation algorithm with approximation ratio ρ for the general TSP.
Theorem 35.3
Proof:Idea: Reduction from the hamiltonian-cycle problem.
Let G = (V ,E) be an instance of the hamiltonian-cycle problem
Let G′ = (V ,E ′) be a complete graph with costs for each (u, v) ∈ E ′:
c(u, v) =
{1 if (u, v) ∈ E ,ρ|V |+ 1 otherwise.
If G has a hamiltonian cycle H, then (G′, c) contains a tour of cost |V |If G does not have a hamiltonian cycle, then any tour T must use some edge 6∈ E ,
⇒ c(T ) ≥ (ρ|V |+ 1) + (|V | − 1)
= (ρ+ 1)|V |.
Gap of ρ+ 1 between tours which are using only edges in G and those which don’tρ-Approximation of TSP in G′ computes hamiltonian cycle in G (if one exists)
Large weight will renderthis edge useless!
Can create representations of G′ andc in time polynomial in |V | and |E |!
G = (V ,E)
Reduction
G′ = (V ,E ′)
1
1
1
1ρ · 4 + 1
1�1ρ · 4 + 1
V. Travelling Salesman Problem General TSP 10
![Page 36: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/36.jpg)
Hardness of Approximation
If P 6= NP, then for any constant ρ ≥ 1, there is no polynomial-time ap-proximation algorithm with approximation ratio ρ for the general TSP.
Theorem 35.3
Proof:Idea: Reduction from the hamiltonian-cycle problem.
Let G = (V ,E) be an instance of the hamiltonian-cycle problemLet G′ = (V ,E ′) be a complete graph with costs for each (u, v) ∈ E ′:
c(u, v) =
{1 if (u, v) ∈ E ,ρ|V |+ 1 otherwise.
If G has a hamiltonian cycle H, then (G′, c) contains a tour of cost |V |If G does not have a hamiltonian cycle, then any tour T must use some edge 6∈ E ,
⇒ c(T ) ≥ (ρ|V |+ 1) + (|V | − 1)
= (ρ+ 1)|V |.
Gap of ρ+ 1 between tours which are using only edges in G and those which don’tρ-Approximation of TSP in G′ computes hamiltonian cycle in G (if one exists)
Large weight will renderthis edge useless!
Can create representations of G′ andc in time polynomial in |V | and |E |!
G = (V ,E)
Reduction
G′ = (V ,E ′)
1
1
1
1ρ · 4 + 1
1�1ρ · 4 + 1
V. Travelling Salesman Problem General TSP 10
![Page 37: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/37.jpg)
Hardness of Approximation
If P 6= NP, then for any constant ρ ≥ 1, there is no polynomial-time ap-proximation algorithm with approximation ratio ρ for the general TSP.
Theorem 35.3
Proof:Idea: Reduction from the hamiltonian-cycle problem.
Let G = (V ,E) be an instance of the hamiltonian-cycle problemLet G′ = (V ,E ′) be a complete graph with costs for each (u, v) ∈ E ′:
c(u, v) =
{1 if (u, v) ∈ E ,ρ|V |+ 1 otherwise.
If G has a hamiltonian cycle H, then (G′, c) contains a tour of cost |V |If G does not have a hamiltonian cycle, then any tour T must use some edge 6∈ E ,
⇒ c(T ) ≥ (ρ|V |+ 1) + (|V | − 1)
= (ρ+ 1)|V |.
Gap of ρ+ 1 between tours which are using only edges in G and those which don’tρ-Approximation of TSP in G′ computes hamiltonian cycle in G (if one exists)
Large weight will renderthis edge useless!
Can create representations of G′ andc in time polynomial in |V | and |E |!
G = (V ,E)
Reduction
G′ = (V ,E ′)
1
1
1
1ρ · 4 + 1
1�1ρ · 4 + 1
V. Travelling Salesman Problem General TSP 10
![Page 38: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/38.jpg)
Hardness of Approximation
If P 6= NP, then for any constant ρ ≥ 1, there is no polynomial-time ap-proximation algorithm with approximation ratio ρ for the general TSP.
Theorem 35.3
Proof:Idea: Reduction from the hamiltonian-cycle problem.
Let G = (V ,E) be an instance of the hamiltonian-cycle problemLet G′ = (V ,E ′) be a complete graph with costs for each (u, v) ∈ E ′:
c(u, v) =
{1 if (u, v) ∈ E ,ρ|V |+ 1 otherwise.
If G has a hamiltonian cycle H, then (G′, c) contains a tour of cost |V |If G does not have a hamiltonian cycle, then any tour T must use some edge 6∈ E ,
⇒ c(T ) ≥ (ρ|V |+ 1) + (|V | − 1)
= (ρ+ 1)|V |.
Gap of ρ+ 1 between tours which are using only edges in G and those which don’tρ-Approximation of TSP in G′ computes hamiltonian cycle in G (if one exists)
Large weight will renderthis edge useless!
Can create representations of G′ andc in time polynomial in |V | and |E |!
G = (V ,E)
Reduction
G′ = (V ,E ′)
1
1
1
1ρ · 4 + 1
1�1ρ · 4 + 1
V. Travelling Salesman Problem General TSP 10
![Page 39: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/39.jpg)
Hardness of Approximation
If P 6= NP, then for any constant ρ ≥ 1, there is no polynomial-time ap-proximation algorithm with approximation ratio ρ for the general TSP.
Theorem 35.3
Proof:Idea: Reduction from the hamiltonian-cycle problem.
Let G = (V ,E) be an instance of the hamiltonian-cycle problemLet G′ = (V ,E ′) be a complete graph with costs for each (u, v) ∈ E ′:
c(u, v) =
{1 if (u, v) ∈ E ,ρ|V |+ 1 otherwise.
If G has a hamiltonian cycle H, then (G′, c) contains a tour of cost |V |If G does not have a hamiltonian cycle, then any tour T must use some edge 6∈ E ,
⇒ c(T ) ≥ (ρ|V |+ 1) + (|V | − 1)
= (ρ+ 1)|V |.
Gap of ρ+ 1 between tours which are using only edges in G and those which don’tρ-Approximation of TSP in G′ computes hamiltonian cycle in G (if one exists)
Large weight will renderthis edge useless!
Can create representations of G′ andc in time polynomial in |V | and |E |!
G = (V ,E)
Reduction
G′ = (V ,E ′)
1
1
1
1ρ · 4 + 1
1
�1ρ · 4 + 1
V. Travelling Salesman Problem General TSP 10
![Page 40: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/40.jpg)
Hardness of Approximation
If P 6= NP, then for any constant ρ ≥ 1, there is no polynomial-time ap-proximation algorithm with approximation ratio ρ for the general TSP.
Theorem 35.3
Proof:Idea: Reduction from the hamiltonian-cycle problem.
Let G = (V ,E) be an instance of the hamiltonian-cycle problemLet G′ = (V ,E ′) be a complete graph with costs for each (u, v) ∈ E ′:
c(u, v) =
{1 if (u, v) ∈ E ,ρ|V |+ 1 otherwise.
If G has a hamiltonian cycle H, then (G′, c) contains a tour of cost |V |If G does not have a hamiltonian cycle, then any tour T must use some edge 6∈ E ,
⇒ c(T ) ≥ (ρ|V |+ 1) + (|V | − 1)
= (ρ+ 1)|V |.
Gap of ρ+ 1 between tours which are using only edges in G and those which don’tρ-Approximation of TSP in G′ computes hamiltonian cycle in G (if one exists)
Large weight will renderthis edge useless!
Can create representations of G′ andc in time polynomial in |V | and |E |!
G = (V ,E)
Reduction
G′ = (V ,E ′)
1
1
1
1ρ · 4 + 1
1
�1ρ · 4 + 1
V. Travelling Salesman Problem General TSP 10
![Page 41: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/41.jpg)
Hardness of Approximation
If P 6= NP, then for any constant ρ ≥ 1, there is no polynomial-time ap-proximation algorithm with approximation ratio ρ for the general TSP.
Theorem 35.3
Proof:Idea: Reduction from the hamiltonian-cycle problem.
Let G = (V ,E) be an instance of the hamiltonian-cycle problemLet G′ = (V ,E ′) be a complete graph with costs for each (u, v) ∈ E ′:
c(u, v) =
{1 if (u, v) ∈ E ,ρ|V |+ 1 otherwise.
If G has a hamiltonian cycle H, then (G′, c) contains a tour of cost |V |If G does not have a hamiltonian cycle, then any tour T must use some edge 6∈ E ,
⇒ c(T ) ≥ (ρ|V |+ 1) + (|V | − 1)
= (ρ+ 1)|V |.
Gap of ρ+ 1 between tours which are using only edges in G and those which don’tρ-Approximation of TSP in G′ computes hamiltonian cycle in G (if one exists)
Large weight will renderthis edge useless!
Can create representations of G′ andc in time polynomial in |V | and |E |!
G = (V ,E)
Reduction
G′ = (V ,E ′)
1
1
1
1ρ · 4 + 1
1
�1ρ · 4 + 1
V. Travelling Salesman Problem General TSP 10
![Page 42: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/42.jpg)
Hardness of Approximation
If P 6= NP, then for any constant ρ ≥ 1, there is no polynomial-time ap-proximation algorithm with approximation ratio ρ for the general TSP.
Theorem 35.3
Proof:Idea: Reduction from the hamiltonian-cycle problem.
Let G = (V ,E) be an instance of the hamiltonian-cycle problemLet G′ = (V ,E ′) be a complete graph with costs for each (u, v) ∈ E ′:
c(u, v) =
{1 if (u, v) ∈ E ,ρ|V |+ 1 otherwise.
If G has a hamiltonian cycle H, then (G′, c) contains a tour of cost |V |If G does not have a hamiltonian cycle, then any tour T must use some edge 6∈ E ,
⇒ c(T ) ≥ (ρ|V |+ 1) + (|V | − 1)
= (ρ+ 1)|V |.
Gap of ρ+ 1 between tours which are using only edges in G and those which don’tρ-Approximation of TSP in G′ computes hamiltonian cycle in G (if one exists)
Large weight will renderthis edge useless!
Can create representations of G′ andc in time polynomial in |V | and |E |!
G = (V ,E)
Reduction
G′ = (V ,E ′)
1
1
1
1ρ · 4 + 1
1
�1ρ · 4 + 1
V. Travelling Salesman Problem General TSP 10
![Page 43: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/43.jpg)
Hardness of Approximation
If P 6= NP, then for any constant ρ ≥ 1, there is no polynomial-time ap-proximation algorithm with approximation ratio ρ for the general TSP.
Theorem 35.3
Proof:Idea: Reduction from the hamiltonian-cycle problem.
Let G = (V ,E) be an instance of the hamiltonian-cycle problemLet G′ = (V ,E ′) be a complete graph with costs for each (u, v) ∈ E ′:
c(u, v) =
{1 if (u, v) ∈ E ,ρ|V |+ 1 otherwise.
If G has a hamiltonian cycle H, then (G′, c) contains a tour of cost |V |
If G does not have a hamiltonian cycle, then any tour T must use some edge 6∈ E ,
⇒ c(T ) ≥ (ρ|V |+ 1) + (|V | − 1)
= (ρ+ 1)|V |.
Gap of ρ+ 1 between tours which are using only edges in G and those which don’tρ-Approximation of TSP in G′ computes hamiltonian cycle in G (if one exists)
Large weight will renderthis edge useless!
Can create representations of G′ andc in time polynomial in |V | and |E |!
G = (V ,E)
Reduction
G′ = (V ,E ′)
1
1
1
1ρ · 4 + 1
1
�1ρ · 4 + 1
V. Travelling Salesman Problem General TSP 10
![Page 44: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/44.jpg)
Hardness of Approximation
If P 6= NP, then for any constant ρ ≥ 1, there is no polynomial-time ap-proximation algorithm with approximation ratio ρ for the general TSP.
Theorem 35.3
Proof:Idea: Reduction from the hamiltonian-cycle problem.
Let G = (V ,E) be an instance of the hamiltonian-cycle problemLet G′ = (V ,E ′) be a complete graph with costs for each (u, v) ∈ E ′:
c(u, v) =
{1 if (u, v) ∈ E ,ρ|V |+ 1 otherwise.
If G has a hamiltonian cycle H, then (G′, c) contains a tour of cost |V |
If G does not have a hamiltonian cycle, then any tour T must use some edge 6∈ E ,
⇒ c(T ) ≥ (ρ|V |+ 1) + (|V | − 1)
= (ρ+ 1)|V |.
Gap of ρ+ 1 between tours which are using only edges in G and those which don’tρ-Approximation of TSP in G′ computes hamiltonian cycle in G (if one exists)
Large weight will renderthis edge useless!
Can create representations of G′ andc in time polynomial in |V | and |E |!
G = (V ,E)
Reduction
G′ = (V ,E ′)
1
1
1
1ρ · 4 + 1
1
�1ρ · 4 + 1
V. Travelling Salesman Problem General TSP 10
![Page 45: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/45.jpg)
Hardness of Approximation
If P 6= NP, then for any constant ρ ≥ 1, there is no polynomial-time ap-proximation algorithm with approximation ratio ρ for the general TSP.
Theorem 35.3
Proof:Idea: Reduction from the hamiltonian-cycle problem.
Let G = (V ,E) be an instance of the hamiltonian-cycle problemLet G′ = (V ,E ′) be a complete graph with costs for each (u, v) ∈ E ′:
c(u, v) =
{1 if (u, v) ∈ E ,ρ|V |+ 1 otherwise.
If G has a hamiltonian cycle H, then (G′, c) contains a tour of cost |V |
If G does not have a hamiltonian cycle, then any tour T must use some edge 6∈ E ,
⇒ c(T ) ≥ (ρ|V |+ 1) + (|V | − 1)
= (ρ+ 1)|V |.
Gap of ρ+ 1 between tours which are using only edges in G and those which don’tρ-Approximation of TSP in G′ computes hamiltonian cycle in G (if one exists)
Large weight will renderthis edge useless!
Can create representations of G′ andc in time polynomial in |V | and |E |!
G = (V ,E)
Reduction
G′ = (V ,E ′)
1
1
1
1ρ · 4 + 1
1
�1ρ · 4 + 1
V. Travelling Salesman Problem General TSP 10
![Page 46: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/46.jpg)
Hardness of Approximation
If P 6= NP, then for any constant ρ ≥ 1, there is no polynomial-time ap-proximation algorithm with approximation ratio ρ for the general TSP.
Theorem 35.3
Proof:Idea: Reduction from the hamiltonian-cycle problem.
Let G = (V ,E) be an instance of the hamiltonian-cycle problemLet G′ = (V ,E ′) be a complete graph with costs for each (u, v) ∈ E ′:
c(u, v) =
{1 if (u, v) ∈ E ,ρ|V |+ 1 otherwise.
If G has a hamiltonian cycle H, then (G′, c) contains a tour of cost |V |
If G does not have a hamiltonian cycle, then any tour T must use some edge 6∈ E ,
⇒ c(T ) ≥ (ρ|V |+ 1) + (|V | − 1)
= (ρ+ 1)|V |.
Gap of ρ+ 1 between tours which are using only edges in G and those which don’tρ-Approximation of TSP in G′ computes hamiltonian cycle in G (if one exists)
Large weight will renderthis edge useless!
Can create representations of G′ andc in time polynomial in |V | and |E |!
G = (V ,E)
Reduction
G′ = (V ,E ′)
1
1
1
1ρ · 4 + 1
1
�1ρ · 4 + 1
V. Travelling Salesman Problem General TSP 10
![Page 47: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/47.jpg)
Hardness of Approximation
If P 6= NP, then for any constant ρ ≥ 1, there is no polynomial-time ap-proximation algorithm with approximation ratio ρ for the general TSP.
Theorem 35.3
Proof:Idea: Reduction from the hamiltonian-cycle problem.
Let G = (V ,E) be an instance of the hamiltonian-cycle problemLet G′ = (V ,E ′) be a complete graph with costs for each (u, v) ∈ E ′:
c(u, v) =
{1 if (u, v) ∈ E ,ρ|V |+ 1 otherwise.
If G has a hamiltonian cycle H, then (G′, c) contains a tour of cost |V |If G does not have a hamiltonian cycle, then any tour T must use some edge 6∈ E ,
⇒ c(T ) ≥ (ρ|V |+ 1) + (|V | − 1)
= (ρ+ 1)|V |.
Gap of ρ+ 1 between tours which are using only edges in G and those which don’tρ-Approximation of TSP in G′ computes hamiltonian cycle in G (if one exists)
Large weight will renderthis edge useless!
Can create representations of G′ andc in time polynomial in |V | and |E |!
G = (V ,E)
Reduction
G′ = (V ,E ′)
1
1
1
1ρ · 4 + 1
1
�1ρ · 4 + 1
V. Travelling Salesman Problem General TSP 10
![Page 48: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/48.jpg)
Hardness of Approximation
If P 6= NP, then for any constant ρ ≥ 1, there is no polynomial-time ap-proximation algorithm with approximation ratio ρ for the general TSP.
Theorem 35.3
Proof:Idea: Reduction from the hamiltonian-cycle problem.
Let G = (V ,E) be an instance of the hamiltonian-cycle problemLet G′ = (V ,E ′) be a complete graph with costs for each (u, v) ∈ E ′:
c(u, v) =
{1 if (u, v) ∈ E ,ρ|V |+ 1 otherwise.
If G has a hamiltonian cycle H, then (G′, c) contains a tour of cost |V |If G does not have a hamiltonian cycle, then any tour T must use some edge 6∈ E ,
⇒ c(T ) ≥ (ρ|V |+ 1) + (|V | − 1)
= (ρ+ 1)|V |.
Gap of ρ+ 1 between tours which are using only edges in G and those which don’tρ-Approximation of TSP in G′ computes hamiltonian cycle in G (if one exists)
Large weight will renderthis edge useless!
Can create representations of G′ andc in time polynomial in |V | and |E |!
G = (V ,E)
Reduction
G′ = (V ,E ′)
1
1
1
1ρ · 4 + 1
1
�1
ρ · 4 + 1
V. Travelling Salesman Problem General TSP 10
![Page 49: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/49.jpg)
Hardness of Approximation
If P 6= NP, then for any constant ρ ≥ 1, there is no polynomial-time ap-proximation algorithm with approximation ratio ρ for the general TSP.
Theorem 35.3
Proof:Idea: Reduction from the hamiltonian-cycle problem.
Let G = (V ,E) be an instance of the hamiltonian-cycle problemLet G′ = (V ,E ′) be a complete graph with costs for each (u, v) ∈ E ′:
c(u, v) =
{1 if (u, v) ∈ E ,ρ|V |+ 1 otherwise.
If G has a hamiltonian cycle H, then (G′, c) contains a tour of cost |V |If G does not have a hamiltonian cycle, then any tour T must use some edge 6∈ E ,
⇒ c(T ) ≥ (ρ|V |+ 1) + (|V | − 1)
= (ρ+ 1)|V |.
Gap of ρ+ 1 between tours which are using only edges in G and those which don’tρ-Approximation of TSP in G′ computes hamiltonian cycle in G (if one exists)
Large weight will renderthis edge useless!
Can create representations of G′ andc in time polynomial in |V | and |E |!
G = (V ,E)
Reduction
G′ = (V ,E ′)
1
1
1
1ρ · 4 + 1
1�1
ρ · 4 + 1
V. Travelling Salesman Problem General TSP 10
![Page 50: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/50.jpg)
Hardness of Approximation
If P 6= NP, then for any constant ρ ≥ 1, there is no polynomial-time ap-proximation algorithm with approximation ratio ρ for the general TSP.
Theorem 35.3
Proof:Idea: Reduction from the hamiltonian-cycle problem.
Let G = (V ,E) be an instance of the hamiltonian-cycle problemLet G′ = (V ,E ′) be a complete graph with costs for each (u, v) ∈ E ′:
c(u, v) =
{1 if (u, v) ∈ E ,ρ|V |+ 1 otherwise.
If G has a hamiltonian cycle H, then (G′, c) contains a tour of cost |V |If G does not have a hamiltonian cycle, then any tour T must use some edge 6∈ E ,
⇒ c(T ) ≥ (ρ|V |+ 1) + (|V | − 1)
= (ρ+ 1)|V |.
Gap of ρ+ 1 between tours which are using only edges in G and those which don’tρ-Approximation of TSP in G′ computes hamiltonian cycle in G (if one exists)
Large weight will renderthis edge useless!
Can create representations of G′ andc in time polynomial in |V | and |E |!
G = (V ,E)
Reduction
G′ = (V ,E ′)
1
1
1
1ρ · 4 + 1
1�1
ρ · 4 + 1
V. Travelling Salesman Problem General TSP 10
![Page 51: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/51.jpg)
Hardness of Approximation
If P 6= NP, then for any constant ρ ≥ 1, there is no polynomial-time ap-proximation algorithm with approximation ratio ρ for the general TSP.
Theorem 35.3
Proof:Idea: Reduction from the hamiltonian-cycle problem.
Let G = (V ,E) be an instance of the hamiltonian-cycle problemLet G′ = (V ,E ′) be a complete graph with costs for each (u, v) ∈ E ′:
c(u, v) =
{1 if (u, v) ∈ E ,ρ|V |+ 1 otherwise.
If G has a hamiltonian cycle H, then (G′, c) contains a tour of cost |V |If G does not have a hamiltonian cycle, then any tour T must use some edge 6∈ E ,
⇒ c(T ) ≥ (ρ|V |+ 1) + (|V | − 1)
= (ρ+ 1)|V |.
Gap of ρ+ 1 between tours which are using only edges in G and those which don’tρ-Approximation of TSP in G′ computes hamiltonian cycle in G (if one exists)
Large weight will renderthis edge useless!
Can create representations of G′ andc in time polynomial in |V | and |E |!
G = (V ,E)
Reduction
G′ = (V ,E ′)
1
1
1
1ρ · 4 + 1
1�1
ρ · 4 + 1
V. Travelling Salesman Problem General TSP 10
![Page 52: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/52.jpg)
Hardness of Approximation
If P 6= NP, then for any constant ρ ≥ 1, there is no polynomial-time ap-proximation algorithm with approximation ratio ρ for the general TSP.
Theorem 35.3
Proof:Idea: Reduction from the hamiltonian-cycle problem.
Let G = (V ,E) be an instance of the hamiltonian-cycle problemLet G′ = (V ,E ′) be a complete graph with costs for each (u, v) ∈ E ′:
c(u, v) =
{1 if (u, v) ∈ E ,ρ|V |+ 1 otherwise.
If G has a hamiltonian cycle H, then (G′, c) contains a tour of cost |V |If G does not have a hamiltonian cycle, then any tour T must use some edge 6∈ E ,
⇒ c(T ) ≥ (ρ|V |+ 1) + (|V | − 1)
= (ρ+ 1)|V |.
Gap of ρ+ 1 between tours which are using only edges in G and those which don’tρ-Approximation of TSP in G′ computes hamiltonian cycle in G (if one exists)
Large weight will renderthis edge useless!
Can create representations of G′ andc in time polynomial in |V | and |E |!
G = (V ,E)
Reduction
G′ = (V ,E ′)
1
1
1
1ρ · 4 + 1
1�1
ρ · 4 + 1
V. Travelling Salesman Problem General TSP 10
![Page 53: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/53.jpg)
Hardness of Approximation
If P 6= NP, then for any constant ρ ≥ 1, there is no polynomial-time ap-proximation algorithm with approximation ratio ρ for the general TSP.
Theorem 35.3
Proof:Idea: Reduction from the hamiltonian-cycle problem.
Let G = (V ,E) be an instance of the hamiltonian-cycle problemLet G′ = (V ,E ′) be a complete graph with costs for each (u, v) ∈ E ′:
c(u, v) =
{1 if (u, v) ∈ E ,ρ|V |+ 1 otherwise.
If G has a hamiltonian cycle H, then (G′, c) contains a tour of cost |V |If G does not have a hamiltonian cycle, then any tour T must use some edge 6∈ E ,
⇒ c(T ) ≥ (ρ|V |+ 1) + (|V | − 1)
= (ρ+ 1)|V |.Gap of ρ+ 1 between tours which are using only edges in G and those which don’tρ-Approximation of TSP in G′ computes hamiltonian cycle in G (if one exists)
Large weight will renderthis edge useless!
Can create representations of G′ andc in time polynomial in |V | and |E |!
G = (V ,E)
Reduction
G′ = (V ,E ′)
1
1
1
1ρ · 4 + 1
1�1
ρ · 4 + 1
V. Travelling Salesman Problem General TSP 10
![Page 54: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/54.jpg)
Hardness of Approximation
If P 6= NP, then for any constant ρ ≥ 1, there is no polynomial-time ap-proximation algorithm with approximation ratio ρ for the general TSP.
Theorem 35.3
Proof:Idea: Reduction from the hamiltonian-cycle problem.
Let G = (V ,E) be an instance of the hamiltonian-cycle problemLet G′ = (V ,E ′) be a complete graph with costs for each (u, v) ∈ E ′:
c(u, v) =
{1 if (u, v) ∈ E ,ρ|V |+ 1 otherwise.
If G has a hamiltonian cycle H, then (G′, c) contains a tour of cost |V |If G does not have a hamiltonian cycle, then any tour T must use some edge 6∈ E ,
⇒ c(T ) ≥ (ρ|V |+ 1) + (|V | − 1) = (ρ+ 1)|V |.
Gap of ρ+ 1 between tours which are using only edges in G and those which don’tρ-Approximation of TSP in G′ computes hamiltonian cycle in G (if one exists)
Large weight will renderthis edge useless!
Can create representations of G′ andc in time polynomial in |V | and |E |!
G = (V ,E)
Reduction
G′ = (V ,E ′)
1
1
1
1ρ · 4 + 1
1�1
ρ · 4 + 1
V. Travelling Salesman Problem General TSP 10
![Page 55: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/55.jpg)
Hardness of Approximation
If P 6= NP, then for any constant ρ ≥ 1, there is no polynomial-time ap-proximation algorithm with approximation ratio ρ for the general TSP.
Theorem 35.3
Proof:Idea: Reduction from the hamiltonian-cycle problem.
Let G = (V ,E) be an instance of the hamiltonian-cycle problemLet G′ = (V ,E ′) be a complete graph with costs for each (u, v) ∈ E ′:
c(u, v) =
{1 if (u, v) ∈ E ,ρ|V |+ 1 otherwise.
If G has a hamiltonian cycle H, then (G′, c) contains a tour of cost |V |If G does not have a hamiltonian cycle, then any tour T must use some edge 6∈ E ,
⇒ c(T ) ≥ (ρ|V |+ 1) + (|V | − 1) = (ρ+ 1)|V |.Gap of ρ+ 1 between tours which are using only edges in G and those which don’t
ρ-Approximation of TSP in G′ computes hamiltonian cycle in G (if one exists)
Large weight will renderthis edge useless!
Can create representations of G′ andc in time polynomial in |V | and |E |!
G = (V ,E)
Reduction
G′ = (V ,E ′)
1
1
1
1ρ · 4 + 1
1�1
ρ · 4 + 1
V. Travelling Salesman Problem General TSP 10
![Page 56: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/56.jpg)
Hardness of Approximation
If P 6= NP, then for any constant ρ ≥ 1, there is no polynomial-time ap-proximation algorithm with approximation ratio ρ for the general TSP.
Theorem 35.3
Proof:Idea: Reduction from the hamiltonian-cycle problem.
Let G = (V ,E) be an instance of the hamiltonian-cycle problemLet G′ = (V ,E ′) be a complete graph with costs for each (u, v) ∈ E ′:
c(u, v) =
{1 if (u, v) ∈ E ,ρ|V |+ 1 otherwise.
If G has a hamiltonian cycle H, then (G′, c) contains a tour of cost |V |If G does not have a hamiltonian cycle, then any tour T must use some edge 6∈ E ,
⇒ c(T ) ≥ (ρ|V |+ 1) + (|V | − 1) = (ρ+ 1)|V |.Gap of ρ+ 1 between tours which are using only edges in G and those which don’tρ-Approximation of TSP in G′ computes hamiltonian cycle in G (if one exists)
Large weight will renderthis edge useless!
Can create representations of G′ andc in time polynomial in |V | and |E |!
G = (V ,E)
Reduction
G′ = (V ,E ′)
1
1
1
1ρ · 4 + 1
1�1
ρ · 4 + 1
V. Travelling Salesman Problem General TSP 10
![Page 57: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/57.jpg)
Hardness of Approximation
If P 6= NP, then for any constant ρ ≥ 1, there is no polynomial-time ap-proximation algorithm with approximation ratio ρ for the general TSP.
Theorem 35.3
Proof:Idea: Reduction from the hamiltonian-cycle problem.
Let G = (V ,E) be an instance of the hamiltonian-cycle problemLet G′ = (V ,E ′) be a complete graph with costs for each (u, v) ∈ E ′:
c(u, v) =
{1 if (u, v) ∈ E ,ρ|V |+ 1 otherwise.
If G has a hamiltonian cycle H, then (G′, c) contains a tour of cost |V |If G does not have a hamiltonian cycle, then any tour T must use some edge 6∈ E ,
⇒ c(T ) ≥ (ρ|V |+ 1) + (|V | − 1) = (ρ+ 1)|V |.Gap of ρ+ 1 between tours which are using only edges in G and those which don’tρ-Approximation of TSP in G′ computes hamiltonian cycle in G (if one exists)
Large weight will renderthis edge useless!
Can create representations of G′ andc in time polynomial in |V | and |E |!
G = (V ,E)
Reduction
G′ = (V ,E ′)
1
1
1
1ρ · 4 + 1
1�1
ρ · 4 + 1
V. Travelling Salesman Problem General TSP 10
![Page 58: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/58.jpg)
Proof of Theorem 35.3 from a higher perspective
x
y
f (x)
f (y)
f
f
instances of Hamilton instances of TSP
All instances with ahamiltonian cycle
All instanceswith cost ≤ k
All instanceswith cost > ρ · k
General Method to prove inapproximability results!
V. Travelling Salesman Problem General TSP 11
![Page 59: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/59.jpg)
Proof of Theorem 35.3 from a higher perspective
x
y
f (x)
f (y)
f
f
instances of Hamilton instances of TSP
All instances with ahamiltonian cycle
All instanceswith cost ≤ k
All instanceswith cost > ρ · k
General Method to prove inapproximability results!
V. Travelling Salesman Problem General TSP 11
![Page 60: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/60.jpg)
Proof of Theorem 35.3 from a higher perspective
x
y
f (x)
f (y)
f
f
instances of Hamilton instances of TSP
All instances with ahamiltonian cycle
All instanceswith cost ≤ k
All instanceswith cost > ρ · k
General Method to prove inapproximability results!
V. Travelling Salesman Problem General TSP 11
![Page 61: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/61.jpg)
Proof of Theorem 35.3 from a higher perspective
x
y
f (x)
f (y)
f
f
instances of Hamilton instances of TSP
All instances with ahamiltonian cycle
All instanceswith cost ≤ k
All instanceswith cost > ρ · k
General Method to prove inapproximability results!
V. Travelling Salesman Problem General TSP 11
![Page 62: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/62.jpg)
Proof of Theorem 35.3 from a higher perspective
x
y
f (x)
f (y)
f
f
instances of Hamilton instances of TSP
All instances with ahamiltonian cycle
All instanceswith cost ≤ k
All instanceswith cost > ρ · k
General Method to prove inapproximability results!
V. Travelling Salesman Problem General TSP 11
![Page 63: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/63.jpg)
Proof of Theorem 35.3 from a higher perspective
x
y
f (x)
f (y)
f
f
instances of Hamilton instances of TSP
All instances with ahamiltonian cycle
All instanceswith cost ≤ k
All instanceswith cost > ρ · k
General Method to prove inapproximability results!
V. Travelling Salesman Problem General TSP 11
![Page 64: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/64.jpg)
Proof of Theorem 35.3 from a higher perspective
x
y
f (x)
f (y)
f
f
instances of Hamilton instances of TSP
All instances with ahamiltonian cycle
All instanceswith cost ≤ k
All instanceswith cost > ρ · k
General Method to prove inapproximability results!
V. Travelling Salesman Problem General TSP 11
![Page 65: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/65.jpg)
Outline
Introduction
General TSP
Metric TSP
V. Travelling Salesman Problem Metric TSP 12
![Page 66: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/66.jpg)
Metric TSP (TSP Problem with the Triangle Inequality)
Idea: First compute an MST, and then create a tour based on the tree.
APPROX-TSP-TOUR(G, c)1: select a vertex r ∈ G.V to be a “root” vertex2: compute a minimum spanning tree Tmin for G from root r3: using MST-PRIM(G, c, r)4: let H be a list of vertices, ordered according to when they are first visited5: in a preorder walk of Tmin
6: return the hamiltonian cycle H
Runtime is dominated by MST-PRIM, which is Θ(V 2).
Remember: In the Metric-TSP problem, G is a complete graph.
V. Travelling Salesman Problem Metric TSP 13
![Page 67: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/67.jpg)
Metric TSP (TSP Problem with the Triangle Inequality)
Idea: First compute an MST, and then create a tour based on the tree.
APPROX-TSP-TOUR(G, c)1: select a vertex r ∈ G.V to be a “root” vertex2: compute a minimum spanning tree Tmin for G from root r3: using MST-PRIM(G, c, r)4: let H be a list of vertices, ordered according to when they are first visited5: in a preorder walk of Tmin
6: return the hamiltonian cycle H
Runtime is dominated by MST-PRIM, which is Θ(V 2).
Remember: In the Metric-TSP problem, G is a complete graph.
V. Travelling Salesman Problem Metric TSP 13
![Page 68: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/68.jpg)
Metric TSP (TSP Problem with the Triangle Inequality)
Idea: First compute an MST, and then create a tour based on the tree.
APPROX-TSP-TOUR(G, c)1: select a vertex r ∈ G.V to be a “root” vertex2: compute a minimum spanning tree Tmin for G from root r3: using MST-PRIM(G, c, r)4: let H be a list of vertices, ordered according to when they are first visited5: in a preorder walk of Tmin
6: return the hamiltonian cycle H
Runtime is dominated by MST-PRIM, which is Θ(V 2).
Remember: In the Metric-TSP problem, G is a complete graph.
V. Travelling Salesman Problem Metric TSP 13
![Page 69: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/69.jpg)
Metric TSP (TSP Problem with the Triangle Inequality)
Idea: First compute an MST, and then create a tour based on the tree.
APPROX-TSP-TOUR(G, c)1: select a vertex r ∈ G.V to be a “root” vertex2: compute a minimum spanning tree Tmin for G from root r3: using MST-PRIM(G, c, r)4: let H be a list of vertices, ordered according to when they are first visited5: in a preorder walk of Tmin
6: return the hamiltonian cycle H
Runtime is dominated by MST-PRIM, which is Θ(V 2).
Remember: In the Metric-TSP problem, G is a complete graph.
V. Travelling Salesman Problem Metric TSP 13
![Page 70: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/70.jpg)
Run of APPROX-TSP-TOUR
a d
b f
e
g
c
h
Solution has cost ≈ 19.704 - not optimal!Better solution, yet still not optimal!This is the optimal solution (cost ≈ 14.715).
1. Compute MST Tmin
X
2. Perform preorder walk on MST Tmin
X
3. Return list of vertices according to the preorder tree walk
X
V. Travelling Salesman Problem Metric TSP 14
![Page 71: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/71.jpg)
Run of APPROX-TSP-TOUR
a d
b f
e
g
c
h
Solution has cost ≈ 19.704 - not optimal!Better solution, yet still not optimal!This is the optimal solution (cost ≈ 14.715).
1. Compute MST Tmin
X
2. Perform preorder walk on MST Tmin
X
3. Return list of vertices according to the preorder tree walk
X
V. Travelling Salesman Problem Metric TSP 14
![Page 72: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/72.jpg)
Run of APPROX-TSP-TOUR
a d
b f
e
g
c
h
Solution has cost ≈ 19.704 - not optimal!Better solution, yet still not optimal!This is the optimal solution (cost ≈ 14.715).
1. Compute MST Tmin
X
2. Perform preorder walk on MST Tmin
X
3. Return list of vertices according to the preorder tree walk
X
V. Travelling Salesman Problem Metric TSP 14
![Page 73: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/73.jpg)
Run of APPROX-TSP-TOUR
a d
b f
e
g
c
h
Solution has cost ≈ 19.704 - not optimal!Better solution, yet still not optimal!This is the optimal solution (cost ≈ 14.715).
1. Compute MST Tmin X
2. Perform preorder walk on MST Tmin
X
3. Return list of vertices according to the preorder tree walk
X
V. Travelling Salesman Problem Metric TSP 14
![Page 74: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/74.jpg)
Run of APPROX-TSP-TOUR
a d
b f
e
g
c
h
Solution has cost ≈ 19.704 - not optimal!Better solution, yet still not optimal!This is the optimal solution (cost ≈ 14.715).
1. Compute MST Tmin X
2. Perform preorder walk on MST Tmin
X
3. Return list of vertices according to the preorder tree walk
X
V. Travelling Salesman Problem Metric TSP 14
![Page 75: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/75.jpg)
Run of APPROX-TSP-TOUR
a d
b f
e
g
c
h
Solution has cost ≈ 19.704 - not optimal!Better solution, yet still not optimal!This is the optimal solution (cost ≈ 14.715).
1. Compute MST Tmin X
2. Perform preorder walk on MST Tmin X
3. Return list of vertices according to the preorder tree walk
X
V. Travelling Salesman Problem Metric TSP 14
![Page 76: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/76.jpg)
Run of APPROX-TSP-TOUR
a d
b f
e
g
c
h
Solution has cost ≈ 19.704 - not optimal!Better solution, yet still not optimal!This is the optimal solution (cost ≈ 14.715).
1. Compute MST Tmin X
2. Perform preorder walk on MST Tmin X
3. Return list of vertices according to the preorder tree walk
X
V. Travelling Salesman Problem Metric TSP 14
![Page 77: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/77.jpg)
Run of APPROX-TSP-TOUR
a d
b f
e
g
c
h
Solution has cost ≈ 19.704 - not optimal!Better solution, yet still not optimal!This is the optimal solution (cost ≈ 14.715).
1. Compute MST Tmin X
2. Perform preorder walk on MST Tmin X
3. Return list of vertices according to the preorder tree walk
X
V. Travelling Salesman Problem Metric TSP 14
![Page 78: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/78.jpg)
Run of APPROX-TSP-TOUR
a d
b f
e
g
c
h
Solution has cost ≈ 19.704 - not optimal!Better solution, yet still not optimal!This is the optimal solution (cost ≈ 14.715).
1. Compute MST Tmin X
2. Perform preorder walk on MST Tmin X
3. Return list of vertices according to the preorder tree walk
X
V. Travelling Salesman Problem Metric TSP 14
![Page 79: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/79.jpg)
Run of APPROX-TSP-TOUR
a d
b f
e
g
c
h
Solution has cost ≈ 19.704 - not optimal!Better solution, yet still not optimal!This is the optimal solution (cost ≈ 14.715).
1. Compute MST Tmin X
2. Perform preorder walk on MST Tmin X
3. Return list of vertices according to the preorder tree walk
X
V. Travelling Salesman Problem Metric TSP 14
![Page 80: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/80.jpg)
Run of APPROX-TSP-TOUR
a d
b f
e
g
c
h
Solution has cost ≈ 19.704 - not optimal!Better solution, yet still not optimal!This is the optimal solution (cost ≈ 14.715).
1. Compute MST Tmin X
2. Perform preorder walk on MST Tmin X
3. Return list of vertices according to the preorder tree walk
X
V. Travelling Salesman Problem Metric TSP 14
![Page 81: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/81.jpg)
Run of APPROX-TSP-TOUR
a d
b f
e
g
c
h
Solution has cost ≈ 19.704 - not optimal!Better solution, yet still not optimal!This is the optimal solution (cost ≈ 14.715).
1. Compute MST Tmin X
2. Perform preorder walk on MST Tmin X
3. Return list of vertices according to the preorder tree walk
X
V. Travelling Salesman Problem Metric TSP 14
![Page 82: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/82.jpg)
Run of APPROX-TSP-TOUR
a d
b f
e
g
c
h
Solution has cost ≈ 19.704 - not optimal!Better solution, yet still not optimal!This is the optimal solution (cost ≈ 14.715).
1. Compute MST Tmin X
2. Perform preorder walk on MST Tmin X
3. Return list of vertices according to the preorder tree walk
X
V. Travelling Salesman Problem Metric TSP 14
![Page 83: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/83.jpg)
Run of APPROX-TSP-TOUR
a d
b f
e
g
c
h
Solution has cost ≈ 19.704 - not optimal!Better solution, yet still not optimal!This is the optimal solution (cost ≈ 14.715).
1. Compute MST Tmin X
2. Perform preorder walk on MST Tmin X
3. Return list of vertices according to the preorder tree walk
X
V. Travelling Salesman Problem Metric TSP 14
![Page 84: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/84.jpg)
Run of APPROX-TSP-TOUR
a d
b f
e
g
c
h
Solution has cost ≈ 19.704 - not optimal!Better solution, yet still not optimal!This is the optimal solution (cost ≈ 14.715).
1. Compute MST Tmin X
2. Perform preorder walk on MST Tmin X
3. Return list of vertices according to the preorder tree walk X
V. Travelling Salesman Problem Metric TSP 14
![Page 85: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/85.jpg)
Run of APPROX-TSP-TOUR
a d
b f
e
g
c
h
Solution has cost ≈ 19.704 - not optimal!
Better solution, yet still not optimal!This is the optimal solution (cost ≈ 14.715).
1. Compute MST Tmin X
2. Perform preorder walk on MST Tmin X
3. Return list of vertices according to the preorder tree walk X
V. Travelling Salesman Problem Metric TSP 14
![Page 86: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/86.jpg)
Run of APPROX-TSP-TOUR
a d
b f
e
g
c
h
Solution has cost ≈ 19.704 - not optimal!Better solution, yet still not optimal!This is the optimal solution (cost ≈ 14.715).
1. Compute MST Tmin X
2. Perform preorder walk on MST Tmin X
3. Return list of vertices according to the preorder tree walk X
V. Travelling Salesman Problem Metric TSP 14
![Page 87: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/87.jpg)
Run of APPROX-TSP-TOUR
a d
b f
e
g
c
h
Solution has cost ≈ 19.704 - not optimal!
Better solution, yet still not optimal!
This is the optimal solution (cost ≈ 14.715).
1. Compute MST Tmin X
2. Perform preorder walk on MST Tmin X
3. Return list of vertices according to the preorder tree walk X
V. Travelling Salesman Problem Metric TSP 14
![Page 88: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/88.jpg)
Run of APPROX-TSP-TOUR
a d
b f
e
g
c
h
Solution has cost ≈ 19.704 - not optimal!Better solution, yet still not optimal!This is the optimal solution (cost ≈ 14.715).
1. Compute MST Tmin X
2. Perform preorder walk on MST Tmin X
3. Return list of vertices according to the preorder tree walk X
V. Travelling Salesman Problem Metric TSP 14
![Page 89: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/89.jpg)
Run of APPROX-TSP-TOUR
a d
b f
e
g
c
h
Solution has cost ≈ 19.704 - not optimal!Better solution, yet still not optimal!
This is the optimal solution (cost ≈ 14.715).
1. Compute MST Tmin X
2. Perform preorder walk on MST Tmin X
3. Return list of vertices according to the preorder tree walk X
V. Travelling Salesman Problem Metric TSP 14
![Page 90: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/90.jpg)
Approximate Solution: Objective 921
1
12
2
3
3
4
4
5
5
6
6
7
7
8
8
9
9
10
10
11
11
12
1213
13
14
14
15
15
16
16
17
17
18
18
19
19
20
20
21
21
22
2223
2324
2425
25
26
2627
27
28
28
29
29
30
3031
31
32
32
33
33
34
3435
35
36
36
37
37
38
38
39
39
40
40
41
41
42
42
11
1
1
1
11
1
1
1
1
111
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
V. Travelling Salesman Problem Metric TSP 15
![Page 91: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/91.jpg)
Optimal Solution: Objective 699
1
12
2
3
3
4
4
5
5
6
6
7
7
8
8
9
9
10
10
11
11
12
1213
13
14
14
15
15
16
16
17
17
18
18
19
19
20
20
21
21
22
2223
2324
2425
25
26
2627
27
28
28
29
29
30
3031
31
32
32
33
33
34
3435
35
36
36
37
37
38
38
39
39
40
40
41
41
42
42
1 1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1 1
1
1
1 1
1
1
1
11
1
1
1
1
1
11
1
1
1
1
1
V. Travelling Salesman Problem Metric TSP 16
![Page 92: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/92.jpg)
Proof of the Approximation Ratio
APPROX-TSP-TOUR is a polynomial-time 2-approximation for thetraveling-salesman problem with the triangle inequality.
Theorem 35.2
Proof:
Consider the optimal tour H∗ and remove an arbitrary edge⇒ yields a spanning tree T and
c(Tmin) ≤
Let W be the full walk of the minimum spanning tree Tmin (including repeated visits)⇒ Full walk traverses every edge exactly twice, so
c(W ) = 2c(Tmin) ≤ 2c(T ) ≤ 2c(H∗)
Deleting duplicate vertices from W yields a tour H
c(H) ≤ c(W ) ≤ 2c(H∗)
exploiting that all edgecosts are non-negative!
exploiting triangle inequality!
a d
b f
e
g
c
h
solution H of APPROX-TSPminimum spanning tree TminWalk W = (a, b, c, b, h, b, a, d, e, f , e, g, e, d, a)Walk W = (a, b, c,�b, h,�b, �a, d, e, f ,�e, g,�e,�d, a)Tour H = (a, b, c, h, d, e, f , g, a)
a d
b f
e
g
c
h
optimal solution H∗spanning tree T as a subset of H∗
V. Travelling Salesman Problem Metric TSP 17
![Page 93: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/93.jpg)
Proof of the Approximation Ratio
APPROX-TSP-TOUR is a polynomial-time 2-approximation for thetraveling-salesman problem with the triangle inequality.
Theorem 35.2
Proof:
Consider the optimal tour H∗ and remove an arbitrary edge⇒ yields a spanning tree T and
c(Tmin) ≤
Let W be the full walk of the minimum spanning tree Tmin (including repeated visits)⇒ Full walk traverses every edge exactly twice, so
c(W ) = 2c(Tmin) ≤ 2c(T ) ≤ 2c(H∗)
Deleting duplicate vertices from W yields a tour H
c(H) ≤ c(W ) ≤ 2c(H∗)
exploiting that all edgecosts are non-negative!
exploiting triangle inequality!
a d
b f
e
g
c
h
solution H of APPROX-TSPminimum spanning tree TminWalk W = (a, b, c, b, h, b, a, d, e, f , e, g, e, d, a)Walk W = (a, b, c,�b, h,�b, �a, d, e, f ,�e, g,�e,�d, a)Tour H = (a, b, c, h, d, e, f , g, a)
a d
b f
e
g
c
h
optimal solution H∗spanning tree T as a subset of H∗
V. Travelling Salesman Problem Metric TSP 17
![Page 94: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/94.jpg)
Proof of the Approximation Ratio
APPROX-TSP-TOUR is a polynomial-time 2-approximation for thetraveling-salesman problem with the triangle inequality.
Theorem 35.2
Proof:
Consider the optimal tour H∗ and remove an arbitrary edge⇒ yields a spanning tree T and
c(Tmin) ≤
Let W be the full walk of the minimum spanning tree Tmin (including repeated visits)⇒ Full walk traverses every edge exactly twice, so
c(W ) = 2c(Tmin) ≤ 2c(T ) ≤ 2c(H∗)
Deleting duplicate vertices from W yields a tour H
c(H) ≤ c(W ) ≤ 2c(H∗)
exploiting that all edgecosts are non-negative!
exploiting triangle inequality!
a d
b f
e
g
c
h
solution H of APPROX-TSP
minimum spanning tree TminWalk W = (a, b, c, b, h, b, a, d, e, f , e, g, e, d, a)Walk W = (a, b, c,�b, h,�b, �a, d, e, f ,�e, g,�e,�d, a)Tour H = (a, b, c, h, d, e, f , g, a)
a d
b f
e
g
c
h
optimal solution H∗spanning tree T as a subset of H∗
V. Travelling Salesman Problem Metric TSP 17
![Page 95: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/95.jpg)
Proof of the Approximation Ratio
APPROX-TSP-TOUR is a polynomial-time 2-approximation for thetraveling-salesman problem with the triangle inequality.
Theorem 35.2
Proof:
Consider the optimal tour H∗ and remove an arbitrary edge⇒ yields a spanning tree T and
c(Tmin) ≤
Let W be the full walk of the minimum spanning tree Tmin (including repeated visits)⇒ Full walk traverses every edge exactly twice, so
c(W ) = 2c(Tmin) ≤ 2c(T ) ≤ 2c(H∗)
Deleting duplicate vertices from W yields a tour H
c(H) ≤ c(W ) ≤ 2c(H∗)
exploiting that all edgecosts are non-negative!
exploiting triangle inequality!
a d
b f
e
g
c
h
solution H of APPROX-TSP
minimum spanning tree TminWalk W = (a, b, c, b, h, b, a, d, e, f , e, g, e, d, a)Walk W = (a, b, c,�b, h,�b, �a, d, e, f ,�e, g,�e,�d, a)Tour H = (a, b, c, h, d, e, f , g, a)
a d
b f
e
g
c
h
optimal solution H∗
spanning tree T as a subset of H∗
V. Travelling Salesman Problem Metric TSP 17
![Page 96: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/96.jpg)
Proof of the Approximation Ratio
APPROX-TSP-TOUR is a polynomial-time 2-approximation for thetraveling-salesman problem with the triangle inequality.
Theorem 35.2
Proof:Consider the optimal tour H∗ and remove an arbitrary edge
⇒ yields a spanning tree T and
c(Tmin) ≤
Let W be the full walk of the minimum spanning tree Tmin (including repeated visits)⇒ Full walk traverses every edge exactly twice, so
c(W ) = 2c(Tmin) ≤ 2c(T ) ≤ 2c(H∗)
Deleting duplicate vertices from W yields a tour H
c(H) ≤ c(W ) ≤ 2c(H∗)
exploiting that all edgecosts are non-negative!
exploiting triangle inequality!
a d
b f
e
g
c
h
solution H of APPROX-TSP
minimum spanning tree TminWalk W = (a, b, c, b, h, b, a, d, e, f , e, g, e, d, a)Walk W = (a, b, c,�b, h,�b, �a, d, e, f ,�e, g,�e,�d, a)Tour H = (a, b, c, h, d, e, f , g, a)
a d
b f
e
g
c
h
optimal solution H∗
spanning tree T as a subset of H∗
V. Travelling Salesman Problem Metric TSP 17
![Page 97: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/97.jpg)
Proof of the Approximation Ratio
APPROX-TSP-TOUR is a polynomial-time 2-approximation for thetraveling-salesman problem with the triangle inequality.
Theorem 35.2
Proof:Consider the optimal tour H∗ and remove an arbitrary edge
⇒ yields a spanning tree T and
c(Tmin) ≤
Let W be the full walk of the minimum spanning tree Tmin (including repeated visits)⇒ Full walk traverses every edge exactly twice, so
c(W ) = 2c(Tmin) ≤ 2c(T ) ≤ 2c(H∗)
Deleting duplicate vertices from W yields a tour H
c(H) ≤ c(W ) ≤ 2c(H∗)
exploiting that all edgecosts are non-negative!
exploiting triangle inequality!
a d
b f
e
g
c
h
solution H of APPROX-TSP
minimum spanning tree TminWalk W = (a, b, c, b, h, b, a, d, e, f , e, g, e, d, a)Walk W = (a, b, c,�b, h,�b, �a, d, e, f ,�e, g,�e,�d, a)Tour H = (a, b, c, h, d, e, f , g, a)
a d
b f
e
g
c
h
optimal solution H∗
spanning tree T as a subset of H∗
V. Travelling Salesman Problem Metric TSP 17
![Page 98: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/98.jpg)
Proof of the Approximation Ratio
APPROX-TSP-TOUR is a polynomial-time 2-approximation for thetraveling-salesman problem with the triangle inequality.
Theorem 35.2
Proof:Consider the optimal tour H∗ and remove an arbitrary edge
⇒ yields a spanning tree T and
c(Tmin) ≤Let W be the full walk of the minimum spanning tree Tmin (including repeated visits)
⇒ Full walk traverses every edge exactly twice, so
c(W ) = 2c(Tmin) ≤ 2c(T ) ≤ 2c(H∗)
Deleting duplicate vertices from W yields a tour H
c(H) ≤ c(W ) ≤ 2c(H∗)
exploiting that all edgecosts are non-negative!
exploiting triangle inequality!
a d
b f
e
g
c
h
solution H of APPROX-TSP
minimum spanning tree TminWalk W = (a, b, c, b, h, b, a, d, e, f , e, g, e, d, a)Walk W = (a, b, c,�b, h,�b, �a, d, e, f ,�e, g,�e,�d, a)Tour H = (a, b, c, h, d, e, f , g, a)
a d
b f
e
g
c
h
optimal solution H∗
spanning tree T as a subset of H∗
V. Travelling Salesman Problem Metric TSP 17
![Page 99: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/99.jpg)
Proof of the Approximation Ratio
APPROX-TSP-TOUR is a polynomial-time 2-approximation for thetraveling-salesman problem with the triangle inequality.
Theorem 35.2
Proof:Consider the optimal tour H∗ and remove an arbitrary edge
⇒ yields a spanning tree T and c(Tmin) ≤ c(T ) ≤ c(H∗)
Let W be the full walk of the minimum spanning tree Tmin (including repeated visits)⇒ Full walk traverses every edge exactly twice, so
c(W ) = 2c(Tmin) ≤ 2c(T ) ≤ 2c(H∗)
Deleting duplicate vertices from W yields a tour H
c(H) ≤ c(W ) ≤ 2c(H∗)
exploiting that all edgecosts are non-negative!
exploiting triangle inequality!
a d
b f
e
g
c
h
solution H of APPROX-TSP
minimum spanning tree TminWalk W = (a, b, c, b, h, b, a, d, e, f , e, g, e, d, a)Walk W = (a, b, c,�b, h,�b, �a, d, e, f ,�e, g,�e,�d, a)Tour H = (a, b, c, h, d, e, f , g, a)
a d
b f
e
g
c
h
optimal solution H∗
spanning tree T as a subset of H∗
V. Travelling Salesman Problem Metric TSP 17
![Page 100: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/100.jpg)
Proof of the Approximation Ratio
APPROX-TSP-TOUR is a polynomial-time 2-approximation for thetraveling-salesman problem with the triangle inequality.
Theorem 35.2
Proof:Consider the optimal tour H∗ and remove an arbitrary edge
⇒ yields a spanning tree T and c(Tmin) ≤ c(T ) ≤ c(H∗)
Let W be the full walk of the minimum spanning tree Tmin (including repeated visits)⇒ Full walk traverses every edge exactly twice, so
c(W ) = 2c(Tmin) ≤ 2c(T ) ≤ 2c(H∗)
Deleting duplicate vertices from W yields a tour H
c(H) ≤ c(W ) ≤ 2c(H∗)
exploiting that all edgecosts are non-negative!
exploiting triangle inequality!
a d
b f
e
g
c
h
solution H of APPROX-TSP
minimum spanning tree TminWalk W = (a, b, c, b, h, b, a, d, e, f , e, g, e, d, a)Walk W = (a, b, c,�b, h,�b, �a, d, e, f ,�e, g,�e,�d, a)Tour H = (a, b, c, h, d, e, f , g, a)
a d
b f
e
g
c
h
optimal solution H∗
spanning tree T as a subset of H∗
V. Travelling Salesman Problem Metric TSP 17
![Page 101: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/101.jpg)
Proof of the Approximation Ratio
APPROX-TSP-TOUR is a polynomial-time 2-approximation for thetraveling-salesman problem with the triangle inequality.
Theorem 35.2
Proof:Consider the optimal tour H∗ and remove an arbitrary edge
⇒ yields a spanning tree T and c(Tmin) ≤ c(T ) ≤ c(H∗)Let W be the full walk of the minimum spanning tree Tmin (including repeated visits)
⇒ Full walk traverses every edge exactly twice, so
c(W ) = 2c(Tmin) ≤ 2c(T ) ≤ 2c(H∗)
Deleting duplicate vertices from W yields a tour H
c(H) ≤ c(W ) ≤ 2c(H∗)
exploiting that all edgecosts are non-negative!
exploiting triangle inequality!
a d
b f
e
g
c
h
solution H of APPROX-TSP
minimum spanning tree TminWalk W = (a, b, c, b, h, b, a, d, e, f , e, g, e, d, a)Walk W = (a, b, c,�b, h,�b, �a, d, e, f ,�e, g,�e,�d, a)Tour H = (a, b, c, h, d, e, f , g, a)
a d
b f
e
g
c
h
optimal solution H∗
spanning tree T as a subset of H∗
V. Travelling Salesman Problem Metric TSP 17
![Page 102: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/102.jpg)
Proof of the Approximation Ratio
APPROX-TSP-TOUR is a polynomial-time 2-approximation for thetraveling-salesman problem with the triangle inequality.
Theorem 35.2
Proof:Consider the optimal tour H∗ and remove an arbitrary edge
⇒ yields a spanning tree T and c(Tmin) ≤ c(T ) ≤ c(H∗)Let W be the full walk of the minimum spanning tree Tmin (including repeated visits)
⇒ Full walk traverses every edge exactly twice, so
c(W ) = 2c(Tmin) ≤ 2c(T ) ≤ 2c(H∗)
Deleting duplicate vertices from W yields a tour H
c(H) ≤ c(W ) ≤ 2c(H∗)
exploiting that all edgecosts are non-negative!
exploiting triangle inequality!
a d
b f
e
g
c
h
solution H of APPROX-TSP
minimum spanning tree Tmin
Walk W = (a, b, c, b, h, b, a, d, e, f , e, g, e, d, a)Walk W = (a, b, c,�b, h,�b, �a, d, e, f ,�e, g,�e,�d, a)Tour H = (a, b, c, h, d, e, f , g, a)
a d
b f
e
g
c
h
optimal solution H∗
spanning tree T as a subset of H∗
V. Travelling Salesman Problem Metric TSP 17
![Page 103: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/103.jpg)
Proof of the Approximation Ratio
APPROX-TSP-TOUR is a polynomial-time 2-approximation for thetraveling-salesman problem with the triangle inequality.
Theorem 35.2
Proof:Consider the optimal tour H∗ and remove an arbitrary edge
⇒ yields a spanning tree T and c(Tmin) ≤ c(T ) ≤ c(H∗)Let W be the full walk of the minimum spanning tree Tmin (including repeated visits)
⇒ Full walk traverses every edge exactly twice, so
c(W ) = 2c(Tmin) ≤ 2c(T ) ≤ 2c(H∗)
Deleting duplicate vertices from W yields a tour H
c(H) ≤ c(W ) ≤ 2c(H∗)
exploiting that all edgecosts are non-negative!
exploiting triangle inequality!
a d
b f
e
g
c
h
solution H of APPROX-TSPminimum spanning tree Tmin
Walk W = (a, b, c, b, h, b, a, d, e, f , e, g, e, d, a)
Walk W = (a, b, c,�b, h,�b, �a, d, e, f ,�e, g,�e,�d, a)Tour H = (a, b, c, h, d, e, f , g, a)
a d
b f
e
g
c
h
optimal solution H∗
spanning tree T as a subset of H∗
V. Travelling Salesman Problem Metric TSP 17
![Page 104: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/104.jpg)
Proof of the Approximation Ratio
APPROX-TSP-TOUR is a polynomial-time 2-approximation for thetraveling-salesman problem with the triangle inequality.
Theorem 35.2
Proof:Consider the optimal tour H∗ and remove an arbitrary edge
⇒ yields a spanning tree T and c(Tmin) ≤ c(T ) ≤ c(H∗)Let W be the full walk of the minimum spanning tree Tmin (including repeated visits)
⇒ Full walk traverses every edge exactly twice, so
c(W ) = 2c(Tmin) ≤ 2c(T ) ≤ 2c(H∗)
Deleting duplicate vertices from W yields a tour H
c(H) ≤ c(W ) ≤ 2c(H∗)
exploiting that all edgecosts are non-negative!
exploiting triangle inequality!
a d
b f
e
g
c
h
solution H of APPROX-TSPminimum spanning tree Tmin
Walk W = (a, b, c, b, h, b, a, d, e, f , e, g, e, d, a)
Walk W = (a, b, c,�b, h,�b, �a, d, e, f ,�e, g,�e,�d, a)Tour H = (a, b, c, h, d, e, f , g, a)
a d
b f
e
g
c
h
optimal solution H∗
spanning tree T as a subset of H∗
V. Travelling Salesman Problem Metric TSP 17
![Page 105: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/105.jpg)
Proof of the Approximation Ratio
APPROX-TSP-TOUR is a polynomial-time 2-approximation for thetraveling-salesman problem with the triangle inequality.
Theorem 35.2
Proof:Consider the optimal tour H∗ and remove an arbitrary edge
⇒ yields a spanning tree T and c(Tmin) ≤ c(T ) ≤ c(H∗)Let W be the full walk of the minimum spanning tree Tmin (including repeated visits)
⇒ Full walk traverses every edge exactly twice, soc(W ) = 2c(Tmin)
≤ 2c(T ) ≤ 2c(H∗)
Deleting duplicate vertices from W yields a tour H
c(H) ≤ c(W ) ≤ 2c(H∗)
exploiting that all edgecosts are non-negative!
exploiting triangle inequality!
a d
b f
e
g
c
h
solution H of APPROX-TSPminimum spanning tree Tmin
Walk W = (a, b, c, b, h, b, a, d, e, f , e, g, e, d, a)
Walk W = (a, b, c,�b, h,�b, �a, d, e, f ,�e, g,�e,�d, a)Tour H = (a, b, c, h, d, e, f , g, a)
a d
b f
e
g
c
h
optimal solution H∗
spanning tree T as a subset of H∗
V. Travelling Salesman Problem Metric TSP 17
![Page 106: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/106.jpg)
Proof of the Approximation Ratio
APPROX-TSP-TOUR is a polynomial-time 2-approximation for thetraveling-salesman problem with the triangle inequality.
Theorem 35.2
Proof:Consider the optimal tour H∗ and remove an arbitrary edge
⇒ yields a spanning tree T and c(Tmin) ≤ c(T ) ≤ c(H∗)Let W be the full walk of the minimum spanning tree Tmin (including repeated visits)
⇒ Full walk traverses every edge exactly twice, soc(W ) = 2c(Tmin) ≤ 2c(T ) ≤ 2c(H∗)
Deleting duplicate vertices from W yields a tour H
c(H) ≤ c(W ) ≤ 2c(H∗)
exploiting that all edgecosts are non-negative!
exploiting triangle inequality!
a d
b f
e
g
c
h
solution H of APPROX-TSPminimum spanning tree Tmin
Walk W = (a, b, c, b, h, b, a, d, e, f , e, g, e, d, a)
Walk W = (a, b, c,�b, h,�b, �a, d, e, f ,�e, g,�e,�d, a)Tour H = (a, b, c, h, d, e, f , g, a)
a d
b f
e
g
c
h
optimal solution H∗
spanning tree T as a subset of H∗
V. Travelling Salesman Problem Metric TSP 17
![Page 107: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/107.jpg)
Proof of the Approximation Ratio
APPROX-TSP-TOUR is a polynomial-time 2-approximation for thetraveling-salesman problem with the triangle inequality.
Theorem 35.2
Proof:Consider the optimal tour H∗ and remove an arbitrary edge
⇒ yields a spanning tree T and c(Tmin) ≤ c(T ) ≤ c(H∗)Let W be the full walk of the minimum spanning tree Tmin (including repeated visits)
⇒ Full walk traverses every edge exactly twice, soc(W ) = 2c(Tmin) ≤ 2c(T ) ≤ 2c(H∗)
Deleting duplicate vertices from W yields a tour H
c(H) ≤ c(W ) ≤ 2c(H∗)
exploiting that all edgecosts are non-negative!
exploiting triangle inequality!
a d
b f
e
g
c
h
solution H of APPROX-TSPminimum spanning tree Tmin
Walk W = (a, b, c, b, h, b, a, d, e, f , e, g, e, d, a)
Walk W = (a, b, c,�b, h,�b, �a, d, e, f ,�e, g,�e,�d, a)Tour H = (a, b, c, h, d, e, f , g, a)
a d
b f
e
g
c
h
optimal solution H∗
spanning tree T as a subset of H∗
V. Travelling Salesman Problem Metric TSP 17
![Page 108: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/108.jpg)
Proof of the Approximation Ratio
APPROX-TSP-TOUR is a polynomial-time 2-approximation for thetraveling-salesman problem with the triangle inequality.
Theorem 35.2
Proof:Consider the optimal tour H∗ and remove an arbitrary edge
⇒ yields a spanning tree T and c(Tmin) ≤ c(T ) ≤ c(H∗)Let W be the full walk of the minimum spanning tree Tmin (including repeated visits)
⇒ Full walk traverses every edge exactly twice, soc(W ) = 2c(Tmin) ≤ 2c(T ) ≤ 2c(H∗)
Deleting duplicate vertices from W yields a tour H
c(H) ≤ c(W ) ≤ 2c(H∗)
exploiting that all edgecosts are non-negative!
exploiting triangle inequality!
a d
b f
e
g
c
h
solution H of APPROX-TSPminimum spanning tree Tmin
Walk W = (a, b, c, b, h, b, a, d, e, f , e, g, e, d, a)
Walk W = (a, b, c,�b, h,�b, �a, d, e, f ,�e, g,�e,�d, a)Tour H = (a, b, c, h, d, e, f , g, a)
a d
b f
e
g
c
h
optimal solution H∗
spanning tree T as a subset of H∗
V. Travelling Salesman Problem Metric TSP 17
![Page 109: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/109.jpg)
Proof of the Approximation Ratio
APPROX-TSP-TOUR is a polynomial-time 2-approximation for thetraveling-salesman problem with the triangle inequality.
Theorem 35.2
Proof:Consider the optimal tour H∗ and remove an arbitrary edge
⇒ yields a spanning tree T and c(Tmin) ≤ c(T ) ≤ c(H∗)Let W be the full walk of the minimum spanning tree Tmin (including repeated visits)
⇒ Full walk traverses every edge exactly twice, soc(W ) = 2c(Tmin) ≤ 2c(T ) ≤ 2c(H∗)
Deleting duplicate vertices from W yields a tour H
c(H) ≤ c(W ) ≤ 2c(H∗)
exploiting that all edgecosts are non-negative!
exploiting triangle inequality!
a d
b f
e
g
c
h
solution H of APPROX-TSPminimum spanning tree TminWalk W = (a, b, c, b, h, b, a, d, e, f , e, g, e, d, a)
Walk W = (a, b, c,�b, h,�b, �a, d, e, f ,�e, g,�e,�d, a)
Tour H = (a, b, c, h, d, e, f , g, a)
a d
b f
e
g
c
h
optimal solution H∗
spanning tree T as a subset of H∗
V. Travelling Salesman Problem Metric TSP 17
![Page 110: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/110.jpg)
Proof of the Approximation Ratio
APPROX-TSP-TOUR is a polynomial-time 2-approximation for thetraveling-salesman problem with the triangle inequality.
Theorem 35.2
Proof:Consider the optimal tour H∗ and remove an arbitrary edge
⇒ yields a spanning tree T and c(Tmin) ≤ c(T ) ≤ c(H∗)Let W be the full walk of the minimum spanning tree Tmin (including repeated visits)
⇒ Full walk traverses every edge exactly twice, soc(W ) = 2c(Tmin) ≤ 2c(T ) ≤ 2c(H∗)
Deleting duplicate vertices from W yields a tour H
c(H) ≤ c(W ) ≤ 2c(H∗)
exploiting that all edgecosts are non-negative!
exploiting triangle inequality!
a d
b f
e
g
c
h
solution H of APPROX-TSPminimum spanning tree TminWalk W = (a, b, c, b, h, b, a, d, e, f , e, g, e, d, a)Walk W = (a, b, c,�b, h,�b, �a, d, e, f ,�e, g,�e,�d, a)
Tour H = (a, b, c, h, d, e, f , g, a)
a d
b f
e
g
c
h
optimal solution H∗
spanning tree T as a subset of H∗
V. Travelling Salesman Problem Metric TSP 17
![Page 111: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/111.jpg)
Proof of the Approximation Ratio
APPROX-TSP-TOUR is a polynomial-time 2-approximation for thetraveling-salesman problem with the triangle inequality.
Theorem 35.2
Proof:Consider the optimal tour H∗ and remove an arbitrary edge
⇒ yields a spanning tree T and c(Tmin) ≤ c(T ) ≤ c(H∗)Let W be the full walk of the minimum spanning tree Tmin (including repeated visits)
⇒ Full walk traverses every edge exactly twice, soc(W ) = 2c(Tmin) ≤ 2c(T ) ≤ 2c(H∗)
Deleting duplicate vertices from W yields a tour H with smaller cost:
c(H) ≤ c(W ) ≤ 2c(H∗)
exploiting that all edgecosts are non-negative!
exploiting triangle inequality!
a d
b f
e
g
c
h
solution H of APPROX-TSPminimum spanning tree TminWalk W = (a, b, c, b, h, b, a, d, e, f , e, g, e, d, a)Walk W = (a, b, c,�b, h,�b, �a, d, e, f ,�e, g,�e,�d, a)
Tour H = (a, b, c, h, d, e, f , g, a)
a d
b f
e
g
c
h
optimal solution H∗
spanning tree T as a subset of H∗
V. Travelling Salesman Problem Metric TSP 17
![Page 112: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/112.jpg)
Proof of the Approximation Ratio
APPROX-TSP-TOUR is a polynomial-time 2-approximation for thetraveling-salesman problem with the triangle inequality.
Theorem 35.2
Proof:Consider the optimal tour H∗ and remove an arbitrary edge
⇒ yields a spanning tree T and c(Tmin) ≤ c(T ) ≤ c(H∗)Let W be the full walk of the minimum spanning tree Tmin (including repeated visits)
⇒ Full walk traverses every edge exactly twice, soc(W ) = 2c(Tmin) ≤ 2c(T ) ≤ 2c(H∗)
Deleting duplicate vertices from W yields a tour H with smaller cost:c(H) ≤ c(W )
≤ 2c(H∗)
exploiting that all edgecosts are non-negative!
exploiting triangle inequality!
a d
b f
e
g
c
h
solution H of APPROX-TSPminimum spanning tree TminWalk W = (a, b, c, b, h, b, a, d, e, f , e, g, e, d, a)Walk W = (a, b, c,�b, h,�b, �a, d, e, f ,�e, g,�e,�d, a)
Tour H = (a, b, c, h, d, e, f , g, a)
a d
b f
e
g
c
h
optimal solution H∗
spanning tree T as a subset of H∗
V. Travelling Salesman Problem Metric TSP 17
![Page 113: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/113.jpg)
Proof of the Approximation Ratio
APPROX-TSP-TOUR is a polynomial-time 2-approximation for thetraveling-salesman problem with the triangle inequality.
Theorem 35.2
Proof:Consider the optimal tour H∗ and remove an arbitrary edge
⇒ yields a spanning tree T and c(Tmin) ≤ c(T ) ≤ c(H∗)Let W be the full walk of the minimum spanning tree Tmin (including repeated visits)
⇒ Full walk traverses every edge exactly twice, soc(W ) = 2c(Tmin) ≤ 2c(T ) ≤ 2c(H∗)
Deleting duplicate vertices from W yields a tour H with smaller cost:c(H) ≤ c(W ) ≤ 2c(H∗)
exploiting that all edgecosts are non-negative!
exploiting triangle inequality!
a d
b f
e
g
c
h
solution H of APPROX-TSPminimum spanning tree TminWalk W = (a, b, c, b, h, b, a, d, e, f , e, g, e, d, a)Walk W = (a, b, c,�b, h,�b, �a, d, e, f ,�e, g,�e,�d, a)
Tour H = (a, b, c, h, d, e, f , g, a)
a d
b f
e
g
c
h
optimal solution H∗
spanning tree T as a subset of H∗
V. Travelling Salesman Problem Metric TSP 17
![Page 114: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/114.jpg)
Proof of the Approximation Ratio
APPROX-TSP-TOUR is a polynomial-time 2-approximation for thetraveling-salesman problem with the triangle inequality.
Theorem 35.2
Proof:Consider the optimal tour H∗ and remove an arbitrary edge
⇒ yields a spanning tree T and c(Tmin) ≤ c(T ) ≤ c(H∗)Let W be the full walk of the minimum spanning tree Tmin (including repeated visits)
⇒ Full walk traverses every edge exactly twice, soc(W ) = 2c(Tmin) ≤ 2c(T ) ≤ 2c(H∗)
Deleting duplicate vertices from W yields a tour H with smaller cost:c(H) ≤ c(W ) ≤ 2c(H∗)
exploiting that all edgecosts are non-negative!
exploiting triangle inequality!
a d
b f
e
g
c
h
solution H of APPROX-TSPminimum spanning tree TminWalk W = (a, b, c, b, h, b, a, d, e, f , e, g, e, d, a)Walk W = (a, b, c,�b, h,�b, �a, d, e, f ,�e, g,�e,�d, a)
Tour H = (a, b, c, h, d, e, f , g, a)
a d
b f
e
g
c
h
optimal solution H∗
spanning tree T as a subset of H∗
V. Travelling Salesman Problem Metric TSP 17
![Page 115: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/115.jpg)
Proof of the Approximation Ratio
APPROX-TSP-TOUR is a polynomial-time 2-approximation for thetraveling-salesman problem with the triangle inequality.
Theorem 35.2
Proof:Consider the optimal tour H∗ and remove an arbitrary edge
⇒ yields a spanning tree T and c(Tmin) ≤ c(T ) ≤ c(H∗)Let W be the full walk of the minimum spanning tree Tmin (including repeated visits)
⇒ Full walk traverses every edge exactly twice, soc(W ) = 2c(Tmin) ≤ 2c(T ) ≤ 2c(H∗)
Deleting duplicate vertices from W yields a tour H with smaller cost:c(H) ≤ c(W ) ≤ 2c(H∗)
exploiting that all edgecosts are non-negative!
exploiting triangle inequality!
a d
b f
e
g
c
h
solution H of APPROX-TSPminimum spanning tree TminWalk W = (a, b, c, b, h, b, a, d, e, f , e, g, e, d, a)Walk W = (a, b, c,�b, h,�b, �a, d, e, f ,�e, g,�e,�d, a)
Tour H = (a, b, c, h, d, e, f , g, a)
a d
b f
e
g
c
h
optimal solution H∗
spanning tree T as a subset of H∗
V. Travelling Salesman Problem Metric TSP 17
![Page 116: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/116.jpg)
Christofides Algorithm
APPROX-TSP-TOUR is a polynomial-time 2-approximation for thetraveling-salesman problem with the triangle inequality.
Theorem 35.2
Can we get a better approximation ratio?
CHRISTOFIDES(G, c)1: select a vertex r ∈ G.V to be a “root” vertex2: compute a minimum spanning tree Tmin for G from root r3: using MST-PRIM(G, c, r)4: compute a perfect matching Mmin with minimum weight in the complete graph5: over the odd-degree vertices in Tmin
6: let H be a list of vertices, ordered according to when they are first visited7: in a Eulearian circuit of Tmin ∪Mmin
8: return the hamiltonian cycle H
There is a polynomial-time 32 -approximation algorithm for the travelling salesman
problem with the triangle inequality.
Theorem (Christofides’76)
V. Travelling Salesman Problem Metric TSP 18
![Page 117: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/117.jpg)
Christofides Algorithm
APPROX-TSP-TOUR is a polynomial-time 2-approximation for thetraveling-salesman problem with the triangle inequality.
Theorem 35.2
Can we get a better approximation ratio?
CHRISTOFIDES(G, c)1: select a vertex r ∈ G.V to be a “root” vertex2: compute a minimum spanning tree Tmin for G from root r3: using MST-PRIM(G, c, r)4: compute a perfect matching Mmin with minimum weight in the complete graph5: over the odd-degree vertices in Tmin
6: let H be a list of vertices, ordered according to when they are first visited7: in a Eulearian circuit of Tmin ∪Mmin
8: return the hamiltonian cycle H
There is a polynomial-time 32 -approximation algorithm for the travelling salesman
problem with the triangle inequality.
Theorem (Christofides’76)
V. Travelling Salesman Problem Metric TSP 18
![Page 118: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/118.jpg)
Christofides Algorithm
APPROX-TSP-TOUR is a polynomial-time 2-approximation for thetraveling-salesman problem with the triangle inequality.
Theorem 35.2
Can we get a better approximation ratio?
CHRISTOFIDES(G, c)1: select a vertex r ∈ G.V to be a “root” vertex2: compute a minimum spanning tree Tmin for G from root r3: using MST-PRIM(G, c, r)4: compute a perfect matching Mmin with minimum weight in the complete graph5: over the odd-degree vertices in Tmin
6: let H be a list of vertices, ordered according to when they are first visited7: in a Eulearian circuit of Tmin ∪Mmin
8: return the hamiltonian cycle H
There is a polynomial-time 32 -approximation algorithm for the travelling salesman
problem with the triangle inequality.
Theorem (Christofides’76)
V. Travelling Salesman Problem Metric TSP 18
![Page 119: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/119.jpg)
Christofides Algorithm
APPROX-TSP-TOUR is a polynomial-time 2-approximation for thetraveling-salesman problem with the triangle inequality.
Theorem 35.2
Can we get a better approximation ratio?
CHRISTOFIDES(G, c)1: select a vertex r ∈ G.V to be a “root” vertex2: compute a minimum spanning tree Tmin for G from root r3: using MST-PRIM(G, c, r)4: compute a perfect matching Mmin with minimum weight in the complete graph5: over the odd-degree vertices in Tmin
6: let H be a list of vertices, ordered according to when they are first visited7: in a Eulearian circuit of Tmin ∪Mmin
8: return the hamiltonian cycle H
There is a polynomial-time 32 -approximation algorithm for the travelling salesman
problem with the triangle inequality.
Theorem (Christofides’76)
V. Travelling Salesman Problem Metric TSP 18
![Page 120: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/120.jpg)
Run of CHRISTOFIDES
a d
b f
e
g
c
h
b f
e
g
c
h
Solution has cost ≈ 15.54 - within 10% of the optimum!
1. Compute MST Tmin
X
2. Add a minimum-weight perfect matching Mmin of the odd vertices in Tmin
X
3. Find an Eulerian Circuit in Tmin ∪Mmin
X
4. Transform the Circuit into a Hamiltonian Cycle
X
All vertices in Tmin ∪Mmin have even degree!
V. Travelling Salesman Problem Metric TSP 19
![Page 121: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/121.jpg)
Run of CHRISTOFIDES
a d
b f
e
g
c
h
b f
e
g
c
h
Solution has cost ≈ 15.54 - within 10% of the optimum!
1. Compute MST Tmin
X
2. Add a minimum-weight perfect matching Mmin of the odd vertices in Tmin
X
3. Find an Eulerian Circuit in Tmin ∪Mmin
X
4. Transform the Circuit into a Hamiltonian Cycle
X
All vertices in Tmin ∪Mmin have even degree!
V. Travelling Salesman Problem Metric TSP 19
![Page 122: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/122.jpg)
Run of CHRISTOFIDES
a d
b f
e
g
c
h
b f
e
g
c
h
Solution has cost ≈ 15.54 - within 10% of the optimum!
1. Compute MST Tmin
X
2. Add a minimum-weight perfect matching Mmin of the odd vertices in Tmin
X
3. Find an Eulerian Circuit in Tmin ∪Mmin
X
4. Transform the Circuit into a Hamiltonian Cycle
X
All vertices in Tmin ∪Mmin have even degree!
V. Travelling Salesman Problem Metric TSP 19
![Page 123: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/123.jpg)
Run of CHRISTOFIDES
a d
b f
e
g
c
h
b f
e
g
c
h
Solution has cost ≈ 15.54 - within 10% of the optimum!
1. Compute MST Tmin X
2. Add a minimum-weight perfect matching Mmin of the odd vertices in Tmin
X
3. Find an Eulerian Circuit in Tmin ∪Mmin
X
4. Transform the Circuit into a Hamiltonian Cycle
X
All vertices in Tmin ∪Mmin have even degree!
V. Travelling Salesman Problem Metric TSP 19
![Page 124: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/124.jpg)
Run of CHRISTOFIDES
a d
b f
e
g
c
h
b f
e
g
c
h
Solution has cost ≈ 15.54 - within 10% of the optimum!
1. Compute MST Tmin X
2. Add a minimum-weight perfect matching Mmin of the odd vertices in Tmin
X
3. Find an Eulerian Circuit in Tmin ∪Mmin
X
4. Transform the Circuit into a Hamiltonian Cycle
X
All vertices in Tmin ∪Mmin have even degree!
V. Travelling Salesman Problem Metric TSP 19
![Page 125: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/125.jpg)
Run of CHRISTOFIDES
a d
b f
e
g
c
h
b f
e
g
c
h
Solution has cost ≈ 15.54 - within 10% of the optimum!
1. Compute MST Tmin X
2. Add a minimum-weight perfect matching Mmin of the odd vertices in Tmin
X
3. Find an Eulerian Circuit in Tmin ∪Mmin
X
4. Transform the Circuit into a Hamiltonian Cycle
X
All vertices in Tmin ∪Mmin have even degree!
V. Travelling Salesman Problem Metric TSP 19
![Page 126: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/126.jpg)
Run of CHRISTOFIDES
a d
b f
e
g
c
h
b f
e
g
c
h
Solution has cost ≈ 15.54 - within 10% of the optimum!
1. Compute MST Tmin X
2. Add a minimum-weight perfect matching Mmin of the odd vertices in Tmin
X
3. Find an Eulerian Circuit in Tmin ∪Mmin
X
4. Transform the Circuit into a Hamiltonian Cycle
X
All vertices in Tmin ∪Mmin have even degree!
V. Travelling Salesman Problem Metric TSP 19
![Page 127: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/127.jpg)
Run of CHRISTOFIDES
a d
b f
e
g
c
h
b f
e
g
c
h
Solution has cost ≈ 15.54 - within 10% of the optimum!
1. Compute MST Tmin X
2. Add a minimum-weight perfect matching Mmin of the odd vertices in Tmin
X
3. Find an Eulerian Circuit in Tmin ∪Mmin
X
4. Transform the Circuit into a Hamiltonian Cycle
X
All vertices in Tmin ∪Mmin have even degree!
V. Travelling Salesman Problem Metric TSP 19
![Page 128: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/128.jpg)
Run of CHRISTOFIDES
a d
b f
e
g
c
h
b f
e
g
c
h
Solution has cost ≈ 15.54 - within 10% of the optimum!
1. Compute MST Tmin X
2. Add a minimum-weight perfect matching Mmin of the odd vertices in Tmin X
3. Find an Eulerian Circuit in Tmin ∪Mmin
X
4. Transform the Circuit into a Hamiltonian Cycle
X
All vertices in Tmin ∪Mmin have even degree!
V. Travelling Salesman Problem Metric TSP 19
![Page 129: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/129.jpg)
Run of CHRISTOFIDES
a d
b f
e
g
c
h
b f
e
g
c
h
Solution has cost ≈ 15.54 - within 10% of the optimum!
1. Compute MST Tmin X
2. Add a minimum-weight perfect matching Mmin of the odd vertices in Tmin X
3. Find an Eulerian Circuit in Tmin ∪Mmin
X
4. Transform the Circuit into a Hamiltonian Cycle
X
All vertices in Tmin ∪Mmin have even degree!
V. Travelling Salesman Problem Metric TSP 19
![Page 130: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/130.jpg)
Run of CHRISTOFIDES
a d
b f
e
g
c
h
b f
e
g
c
h
Solution has cost ≈ 15.54 - within 10% of the optimum!
1. Compute MST Tmin X
2. Add a minimum-weight perfect matching Mmin of the odd vertices in Tmin X
3. Find an Eulerian Circuit in Tmin ∪Mmin X
4. Transform the Circuit into a Hamiltonian Cycle
X
All vertices in Tmin ∪Mmin have even degree!
V. Travelling Salesman Problem Metric TSP 19
![Page 131: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/131.jpg)
Run of CHRISTOFIDES
a d
b f
e
g
c
h
b f
e
g
c
h
Solution has cost ≈ 15.54 - within 10% of the optimum!
1. Compute MST Tmin X
2. Add a minimum-weight perfect matching Mmin of the odd vertices in Tmin X
3. Find an Eulerian Circuit in Tmin ∪Mmin X
4. Transform the Circuit into a Hamiltonian Cycle
XAll vertices in Tmin ∪Mmin have even degree!
V. Travelling Salesman Problem Metric TSP 19
![Page 132: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/132.jpg)
Run of CHRISTOFIDES
a d
b f
e
g
c
h
b f
e
g
c
h
Solution has cost ≈ 15.54 - within 10% of the optimum!
1. Compute MST Tmin X
2. Add a minimum-weight perfect matching Mmin of the odd vertices in Tmin X
3. Find an Eulerian Circuit in Tmin ∪Mmin X
4. Transform the Circuit into a Hamiltonian Cycle
XAll vertices in Tmin ∪Mmin have even degree!
V. Travelling Salesman Problem Metric TSP 19
![Page 133: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/133.jpg)
Run of CHRISTOFIDES
a d
b f
e
g
c
h
b f
e
g
c
h
Solution has cost ≈ 15.54 - within 10% of the optimum!
1. Compute MST Tmin X
2. Add a minimum-weight perfect matching Mmin of the odd vertices in Tmin X
3. Find an Eulerian Circuit in Tmin ∪Mmin X
4. Transform the Circuit into a Hamiltonian Cycle
XAll vertices in Tmin ∪Mmin have even degree!
V. Travelling Salesman Problem Metric TSP 19
![Page 134: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/134.jpg)
Run of CHRISTOFIDES
a d
b f
e
g
c
h
b f
e
g
c
h
Solution has cost ≈ 15.54 - within 10% of the optimum!
1. Compute MST Tmin X
2. Add a minimum-weight perfect matching Mmin of the odd vertices in Tmin X
3. Find an Eulerian Circuit in Tmin ∪Mmin X
4. Transform the Circuit into a Hamiltonian Cycle
XAll vertices in Tmin ∪Mmin have even degree!
V. Travelling Salesman Problem Metric TSP 19
![Page 135: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/135.jpg)
Run of CHRISTOFIDES
a d
b f
e
g
c
h
b f
e
g
c
h
Solution has cost ≈ 15.54 - within 10% of the optimum!
1. Compute MST Tmin X
2. Add a minimum-weight perfect matching Mmin of the odd vertices in Tmin X
3. Find an Eulerian Circuit in Tmin ∪Mmin X
4. Transform the Circuit into a Hamiltonian Cycle
XAll vertices in Tmin ∪Mmin have even degree!
V. Travelling Salesman Problem Metric TSP 19
![Page 136: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/136.jpg)
Run of CHRISTOFIDES
a d
b f
e
g
c
h
b f
e
g
c
h
Solution has cost ≈ 15.54 - within 10% of the optimum!
1. Compute MST Tmin X
2. Add a minimum-weight perfect matching Mmin of the odd vertices in Tmin X
3. Find an Eulerian Circuit in Tmin ∪Mmin X
4. Transform the Circuit into a Hamiltonian Cycle
XAll vertices in Tmin ∪Mmin have even degree!
V. Travelling Salesman Problem Metric TSP 19
![Page 137: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/137.jpg)
Run of CHRISTOFIDES
a d
b f
e
g
c
h
b f
e
g
c
h
Solution has cost ≈ 15.54 - within 10% of the optimum!
1. Compute MST Tmin X
2. Add a minimum-weight perfect matching Mmin of the odd vertices in Tmin X
3. Find an Eulerian Circuit in Tmin ∪Mmin X
4. Transform the Circuit into a Hamiltonian Cycle
XAll vertices in Tmin ∪Mmin have even degree!
V. Travelling Salesman Problem Metric TSP 19
![Page 138: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/138.jpg)
Run of CHRISTOFIDES
a d
b f
e
g
c
h
b f
e
g
c
h
Solution has cost ≈ 15.54 - within 10% of the optimum!
1. Compute MST Tmin X
2. Add a minimum-weight perfect matching Mmin of the odd vertices in Tmin X
3. Find an Eulerian Circuit in Tmin ∪Mmin X
4. Transform the Circuit into a Hamiltonian Cycle
XAll vertices in Tmin ∪Mmin have even degree!
V. Travelling Salesman Problem Metric TSP 19
![Page 139: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/139.jpg)
Run of CHRISTOFIDES
a d
b f
e
g
c
h
b f
e
g
c
h
Solution has cost ≈ 15.54 - within 10% of the optimum!
1. Compute MST Tmin X
2. Add a minimum-weight perfect matching Mmin of the odd vertices in Tmin X
3. Find an Eulerian Circuit in Tmin ∪Mmin X
4. Transform the Circuit into a Hamiltonian Cycle
XAll vertices in Tmin ∪Mmin have even degree!
V. Travelling Salesman Problem Metric TSP 19
![Page 140: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/140.jpg)
Run of CHRISTOFIDES
a d
b f
e
g
c
h
b f
e
g
c
h
Solution has cost ≈ 15.54 - within 10% of the optimum!
1. Compute MST Tmin X
2. Add a minimum-weight perfect matching Mmin of the odd vertices in Tmin X
3. Find an Eulerian Circuit in Tmin ∪Mmin X
4. Transform the Circuit into a Hamiltonian Cycle
XAll vertices in Tmin ∪Mmin have even degree!
V. Travelling Salesman Problem Metric TSP 19
![Page 141: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/141.jpg)
Run of CHRISTOFIDES
a d
b f
e
g
c
h
b f
e
g
c
h
Solution has cost ≈ 15.54 - within 10% of the optimum!
1. Compute MST Tmin X
2. Add a minimum-weight perfect matching Mmin of the odd vertices in Tmin X
3. Find an Eulerian Circuit in Tmin ∪Mmin X
4. Transform the Circuit into a Hamiltonian Cycle X
All vertices in Tmin ∪Mmin have even degree!
V. Travelling Salesman Problem Metric TSP 19
![Page 142: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/142.jpg)
Run of CHRISTOFIDES
a d
b f
e
g
c
h
b f
e
g
c
h
Solution has cost ≈ 15.54 - within 10% of the optimum!
1. Compute MST Tmin X
2. Add a minimum-weight perfect matching Mmin of the odd vertices in Tmin X
3. Find an Eulerian Circuit in Tmin ∪Mmin X
4. Transform the Circuit into a Hamiltonian Cycle X
All vertices in Tmin ∪Mmin have even degree!
V. Travelling Salesman Problem Metric TSP 19
![Page 143: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/143.jpg)
Proof of the Approximation Ratio
There is a polynomial-time 32 -approximation algorithm for the travelling
salesman problem with the triangle inequality.
Theorem (Christofides’76)
Proof is quite similar to the previous analysisProof (Approximation Ratio):
As before, let H∗ denote the optimal tourThe Eulerian Circuit W uses each edge of the minimum spanning treeTmin and the minimum-weight matching Mmin exactly once:
c(W ) = c(Tmin) + c(Mmin) ≤ c(H∗) + c(Mmin)
(1)
Let H∗odd be an optimal tour on the odd-degree vertices in Tmin
Taking edges alternately, we obtain two matchings M1 and M2 such thatc(M1) + c(M2) = c(H∗odd )By shortcutting and the triangle inequality,
c(Mmin) ≤ 12
c(Hodd∗) ≤ 1
2c(H∗).
(2)
Combining 1 with 2 yields
c(W ) ≤ c(H∗) + c(Mmin) ≤ c(H∗) +12
c(H∗) =32
c(H∗).
Number of odd-degree vertices is even!
V. Travelling Salesman Problem Metric TSP 20
![Page 144: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/144.jpg)
Proof of the Approximation Ratio
There is a polynomial-time 32 -approximation algorithm for the travelling
salesman problem with the triangle inequality.
Theorem (Christofides’76)
Proof is quite similar to the previous analysisProof (Approximation Ratio):
As before, let H∗ denote the optimal tourThe Eulerian Circuit W uses each edge of the minimum spanning treeTmin and the minimum-weight matching Mmin exactly once:
c(W ) = c(Tmin) + c(Mmin) ≤ c(H∗) + c(Mmin)
(1)
Let H∗odd be an optimal tour on the odd-degree vertices in Tmin
Taking edges alternately, we obtain two matchings M1 and M2 such thatc(M1) + c(M2) = c(H∗odd )By shortcutting and the triangle inequality,
c(Mmin) ≤ 12
c(Hodd∗) ≤ 1
2c(H∗).
(2)
Combining 1 with 2 yields
c(W ) ≤ c(H∗) + c(Mmin) ≤ c(H∗) +12
c(H∗) =32
c(H∗).
Number of odd-degree vertices is even!
V. Travelling Salesman Problem Metric TSP 20
![Page 145: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/145.jpg)
Proof of the Approximation Ratio
There is a polynomial-time 32 -approximation algorithm for the travelling
salesman problem with the triangle inequality.
Theorem (Christofides’76)
Proof is quite similar to the previous analysisProof (Approximation Ratio):
As before, let H∗ denote the optimal tour
The Eulerian Circuit W uses each edge of the minimum spanning treeTmin and the minimum-weight matching Mmin exactly once:
c(W ) = c(Tmin) + c(Mmin) ≤ c(H∗) + c(Mmin)
(1)
Let H∗odd be an optimal tour on the odd-degree vertices in Tmin
Taking edges alternately, we obtain two matchings M1 and M2 such thatc(M1) + c(M2) = c(H∗odd )By shortcutting and the triangle inequality,
c(Mmin) ≤ 12
c(Hodd∗) ≤ 1
2c(H∗).
(2)
Combining 1 with 2 yields
c(W ) ≤ c(H∗) + c(Mmin) ≤ c(H∗) +12
c(H∗) =32
c(H∗).
Number of odd-degree vertices is even!
V. Travelling Salesman Problem Metric TSP 20
![Page 146: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/146.jpg)
Proof of the Approximation Ratio
There is a polynomial-time 32 -approximation algorithm for the travelling
salesman problem with the triangle inequality.
Theorem (Christofides’76)
Proof is quite similar to the previous analysisProof (Approximation Ratio):
As before, let H∗ denote the optimal tourThe Eulerian Circuit W uses each edge of the minimum spanning treeTmin and the minimum-weight matching Mmin exactly once:
c(W ) = c(Tmin) + c(Mmin) ≤ c(H∗) + c(Mmin)
(1)
Let H∗odd be an optimal tour on the odd-degree vertices in Tmin
Taking edges alternately, we obtain two matchings M1 and M2 such thatc(M1) + c(M2) = c(H∗odd )By shortcutting and the triangle inequality,
c(Mmin) ≤ 12
c(Hodd∗) ≤ 1
2c(H∗).
(2)
Combining 1 with 2 yields
c(W ) ≤ c(H∗) + c(Mmin) ≤ c(H∗) +12
c(H∗) =32
c(H∗).
Number of odd-degree vertices is even!
V. Travelling Salesman Problem Metric TSP 20
![Page 147: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/147.jpg)
Proof of the Approximation Ratio
There is a polynomial-time 32 -approximation algorithm for the travelling
salesman problem with the triangle inequality.
Theorem (Christofides’76)
Proof is quite similar to the previous analysisProof (Approximation Ratio):
As before, let H∗ denote the optimal tourThe Eulerian Circuit W uses each edge of the minimum spanning treeTmin and the minimum-weight matching Mmin exactly once:
c(W ) = c(Tmin) + c(Mmin) ≤ c(H∗) + c(Mmin) (1)
Let H∗odd be an optimal tour on the odd-degree vertices in Tmin
Taking edges alternately, we obtain two matchings M1 and M2 such thatc(M1) + c(M2) = c(H∗odd )By shortcutting and the triangle inequality,
c(Mmin) ≤ 12
c(Hodd∗) ≤ 1
2c(H∗).
(2)
Combining 1 with 2 yields
c(W ) ≤ c(H∗) + c(Mmin) ≤ c(H∗) +12
c(H∗) =32
c(H∗).
Number of odd-degree vertices is even!
V. Travelling Salesman Problem Metric TSP 20
![Page 148: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/148.jpg)
Proof of the Approximation Ratio
There is a polynomial-time 32 -approximation algorithm for the travelling
salesman problem with the triangle inequality.
Theorem (Christofides’76)
Proof is quite similar to the previous analysisProof (Approximation Ratio):
As before, let H∗ denote the optimal tourThe Eulerian Circuit W uses each edge of the minimum spanning treeTmin and the minimum-weight matching Mmin exactly once:
c(W ) = c(Tmin) + c(Mmin) ≤ c(H∗) + c(Mmin) (1)
Let H∗odd be an optimal tour on the odd-degree vertices in Tmin
Taking edges alternately, we obtain two matchings M1 and M2 such thatc(M1) + c(M2) = c(H∗odd )By shortcutting and the triangle inequality,
c(Mmin) ≤ 12
c(Hodd∗) ≤ 1
2c(H∗).
(2)
Combining 1 with 2 yields
c(W ) ≤ c(H∗) + c(Mmin) ≤ c(H∗) +12
c(H∗) =32
c(H∗).
Number of odd-degree vertices is even!
V. Travelling Salesman Problem Metric TSP 20
![Page 149: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/149.jpg)
Proof of the Approximation Ratio
There is a polynomial-time 32 -approximation algorithm for the travelling
salesman problem with the triangle inequality.
Theorem (Christofides’76)
Proof is quite similar to the previous analysisProof (Approximation Ratio):
As before, let H∗ denote the optimal tourThe Eulerian Circuit W uses each edge of the minimum spanning treeTmin and the minimum-weight matching Mmin exactly once:
c(W ) = c(Tmin) + c(Mmin) ≤ c(H∗) + c(Mmin) (1)
Let H∗odd be an optimal tour on the odd-degree vertices in Tmin
Taking edges alternately, we obtain two matchings M1 and M2 such thatc(M1) + c(M2) = c(H∗odd )
By shortcutting and the triangle inequality,
c(Mmin) ≤ 12
c(Hodd∗) ≤ 1
2c(H∗).
(2)
Combining 1 with 2 yields
c(W ) ≤ c(H∗) + c(Mmin) ≤ c(H∗) +12
c(H∗) =32
c(H∗).
Number of odd-degree vertices is even!
V. Travelling Salesman Problem Metric TSP 20
![Page 150: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/150.jpg)
Proof of the Approximation Ratio
There is a polynomial-time 32 -approximation algorithm for the travelling
salesman problem with the triangle inequality.
Theorem (Christofides’76)
Proof is quite similar to the previous analysisProof (Approximation Ratio):
As before, let H∗ denote the optimal tourThe Eulerian Circuit W uses each edge of the minimum spanning treeTmin and the minimum-weight matching Mmin exactly once:
c(W ) = c(Tmin) + c(Mmin) ≤ c(H∗) + c(Mmin) (1)
Let H∗odd be an optimal tour on the odd-degree vertices in Tmin
Taking edges alternately, we obtain two matchings M1 and M2 such thatc(M1) + c(M2) = c(H∗odd )
By shortcutting and the triangle inequality,
c(Mmin) ≤ 12
c(Hodd∗) ≤ 1
2c(H∗).
(2)
Combining 1 with 2 yields
c(W ) ≤ c(H∗) + c(Mmin) ≤ c(H∗) +12
c(H∗) =32
c(H∗).
Number of odd-degree vertices is even!
V. Travelling Salesman Problem Metric TSP 20
![Page 151: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/151.jpg)
Proof of the Approximation Ratio
There is a polynomial-time 32 -approximation algorithm for the travelling
salesman problem with the triangle inequality.
Theorem (Christofides’76)
Proof is quite similar to the previous analysisProof (Approximation Ratio):
As before, let H∗ denote the optimal tourThe Eulerian Circuit W uses each edge of the minimum spanning treeTmin and the minimum-weight matching Mmin exactly once:
c(W ) = c(Tmin) + c(Mmin) ≤ c(H∗) + c(Mmin) (1)
Let H∗odd be an optimal tour on the odd-degree vertices in Tmin
Taking edges alternately, we obtain two matchings M1 and M2 such thatc(M1) + c(M2) = c(H∗odd )By shortcutting and the triangle inequality,
c(Mmin) ≤ 12
c(Hodd∗) ≤ 1
2c(H∗).
(2)
Combining 1 with 2 yields
c(W ) ≤ c(H∗) + c(Mmin) ≤ c(H∗) +12
c(H∗) =32
c(H∗).
Number of odd-degree vertices is even!
V. Travelling Salesman Problem Metric TSP 20
![Page 152: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/152.jpg)
Proof of the Approximation Ratio
There is a polynomial-time 32 -approximation algorithm for the travelling
salesman problem with the triangle inequality.
Theorem (Christofides’76)
Proof is quite similar to the previous analysisProof (Approximation Ratio):
As before, let H∗ denote the optimal tourThe Eulerian Circuit W uses each edge of the minimum spanning treeTmin and the minimum-weight matching Mmin exactly once:
c(W ) = c(Tmin) + c(Mmin) ≤ c(H∗) + c(Mmin) (1)
Let H∗odd be an optimal tour on the odd-degree vertices in Tmin
Taking edges alternately, we obtain two matchings M1 and M2 such thatc(M1) + c(M2) = c(H∗odd )By shortcutting and the triangle inequality,
c(Mmin) ≤ 12
c(Hodd∗) ≤ 1
2c(H∗). (2)
Combining 1 with 2 yields
c(W ) ≤ c(H∗) + c(Mmin) ≤ c(H∗) +12
c(H∗) =32
c(H∗).
Number of odd-degree vertices is even!
V. Travelling Salesman Problem Metric TSP 20
![Page 153: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/153.jpg)
Proof of the Approximation Ratio
There is a polynomial-time 32 -approximation algorithm for the travelling
salesman problem with the triangle inequality.
Theorem (Christofides’76)
Proof is quite similar to the previous analysisProof (Approximation Ratio):
As before, let H∗ denote the optimal tourThe Eulerian Circuit W uses each edge of the minimum spanning treeTmin and the minimum-weight matching Mmin exactly once:
c(W ) = c(Tmin) + c(Mmin) ≤ c(H∗) + c(Mmin) (1)
Let H∗odd be an optimal tour on the odd-degree vertices in Tmin
Taking edges alternately, we obtain two matchings M1 and M2 such thatc(M1) + c(M2) = c(H∗odd )By shortcutting and the triangle inequality,
c(Mmin) ≤ 12
c(Hodd∗) ≤ 1
2c(H∗). (2)
Combining 1 with 2 yields
c(W ) ≤ c(H∗) + c(Mmin) ≤ c(H∗) +12
c(H∗) =32
c(H∗).
Number of odd-degree vertices is even!
V. Travelling Salesman Problem Metric TSP 20
![Page 154: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/154.jpg)
Proof of the Approximation Ratio
There is a polynomial-time 32 -approximation algorithm for the travelling
salesman problem with the triangle inequality.
Theorem (Christofides’76)
Proof is quite similar to the previous analysisProof (Approximation Ratio):
As before, let H∗ denote the optimal tourThe Eulerian Circuit W uses each edge of the minimum spanning treeTmin and the minimum-weight matching Mmin exactly once:
c(W ) = c(Tmin) + c(Mmin) ≤ c(H∗) + c(Mmin) (1)
Let H∗odd be an optimal tour on the odd-degree vertices in Tmin
Taking edges alternately, we obtain two matchings M1 and M2 such thatc(M1) + c(M2) = c(H∗odd )By shortcutting and the triangle inequality,
c(Mmin) ≤ 12
c(Hodd∗) ≤ 1
2c(H∗). (2)
Combining 1 with 2 yields
c(W ) ≤ c(H∗) + c(Mmin) ≤ c(H∗) +12
c(H∗) =32
c(H∗).
Number of odd-degree vertices is even!
V. Travelling Salesman Problem Metric TSP 20
![Page 155: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/155.jpg)
Concluding Remarks
There is a polynomial-time 32 -approximation algorithm for the travelling
salesman problem with the triangle inequality.
Theorem (Christofides’76)
still the best algorithm for the metric TSP problem(!)
There is a PTAS for the Euclidean TSP Problem.Theorem (Arora’96, Mitchell’96)
Both received the Gödel Award 2010
“Christos Papadimitriou told me that the traveling sales-man problem is not a problem. It’s an addiction.”
Jon Bentley 1991
V. Travelling Salesman Problem Metric TSP 21
![Page 156: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/156.jpg)
Concluding Remarks
There is a polynomial-time 32 -approximation algorithm for the travelling
salesman problem with the triangle inequality.
Theorem (Christofides’76)
still the best algorithm for the metric TSP problem(!)
There is a PTAS for the Euclidean TSP Problem.Theorem (Arora’96, Mitchell’96)
Both received the Gödel Award 2010
“Christos Papadimitriou told me that the traveling sales-man problem is not a problem. It’s an addiction.”
Jon Bentley 1991
V. Travelling Salesman Problem Metric TSP 21
![Page 157: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/157.jpg)
Concluding Remarks
There is a polynomial-time 32 -approximation algorithm for the travelling
salesman problem with the triangle inequality.
Theorem (Christofides’76)
still the best algorithm for the metric TSP problem(!)
There is a PTAS for the Euclidean TSP Problem.Theorem (Arora’96, Mitchell’96)
Both received the Gödel Award 2010
“Christos Papadimitriou told me that the traveling sales-man problem is not a problem. It’s an addiction.”
Jon Bentley 1991
V. Travelling Salesman Problem Metric TSP 21
![Page 158: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/158.jpg)
Concluding Remarks
There is a polynomial-time 32 -approximation algorithm for the travelling
salesman problem with the triangle inequality.
Theorem (Christofides’76)
still the best algorithm for the metric TSP problem(!)
There is a PTAS for the Euclidean TSP Problem.Theorem (Arora’96, Mitchell’96)
Both received the Gödel Award 2010
“Christos Papadimitriou told me that the traveling sales-man problem is not a problem. It’s an addiction.”
Jon Bentley 1991
V. Travelling Salesman Problem Metric TSP 21
![Page 159: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/159.jpg)
Concluding Remarks
There is a polynomial-time 32 -approximation algorithm for the travelling
salesman problem with the triangle inequality.
Theorem (Christofides’76)
still the best algorithm for the metric TSP problem(!)
There is a PTAS for the Euclidean TSP Problem.Theorem (Arora’96, Mitchell’96)
Both received the Gödel Award 2010
“Christos Papadimitriou told me that the traveling sales-man problem is not a problem. It’s an addiction.”
Jon Bentley 1991
V. Travelling Salesman Problem Metric TSP 21
![Page 160: V. Approx. Algorithms: Travelling Salesman Problem · 2020. 4. 23. · The traveling salesman problem recently achieved national prominence when a soap company used it as the basis](https://reader033.vdocuments.us/reader033/viewer/2022060909/60a45702511346111d7f6858/html5/thumbnails/160.jpg)
Exercise: Prove that the approximation ratio of APPROX-TSP-TOURsatisfies ρ(n)<2.Hint: Consider the effect of the shortcutting, but note that edge costsmight be zero!
V. Travelling Salesman Problem Metric TSP 22