evolving exact algorithms with gp - hfuu.edu.cn
TRANSCRIPT
![Page 1: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/1.jpg)
Evolving Exact Integer Algorithms withGenetic Programming
Thomas Weise, Mingxu Wan, Ke Tang, and Xin YaoNature Inspired Computation and Applications Laboratory
University of Science and Technology of China (USTC)
Hefei 230027, Anhui, China
[email protected] · http://www.it-weise.de/
IEEE Congress on Evolutionary Computation (IEEE CEC’14) @
2014 IEEE World Congress on Computational Intelligence (WCCI’14)
2014-07-10 ThE1-3 13:50 at Beijing, China,
Beijing International Convention Center, Room 203C
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 1/27
![Page 2: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/2.jpg)
Contents
This is the presentation for conference paper [? ]. You can find the citation
information and reference at the end of these slides, including a download
link for the paper.
1 Benchmark Problems
2 Investigated Ideas
3 Experiments
4 Summary
5 Evolved Programs
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 2/27
![Page 3: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/3.jpg)
Introduction
Evolve algorithms with Genetic Programming.
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 3/27
![Page 4: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/4.jpg)
Introduction
Evolve algorithms with Genetic Programming.
Algorithms ≡ instructions
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 3/27
![Page 5: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/5.jpg)
Introduction
Evolve algorithms with Genetic Programming.
Algorithms ≡ instructions + control flow
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 3/27
![Page 6: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/6.jpg)
Introduction
Evolve algorithms with Genetic Programming.
Algorithms ≡ instructions + control flow + memory
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 3/27
![Page 7: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/7.jpg)
Introduction
Evolve algorithms with Genetic Programming.
Algorithms ≡ instructions + control flow + memory 6= Formula
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 3/27
![Page 8: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/8.jpg)
Introduction
Evolve algorithms with Genetic Programming.
Algorithms ≡ instructions + control flow + memory 6= Formula
Exact integer computations
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 3/27
![Page 9: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/9.jpg)
Introduction
Evolve algorithms with Genetic Programming.
Algorithms ≡ instructions + control flow + memory 6= Formula
Exact integer computations: Result is either right or wrong
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 3/27
![Page 10: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/10.jpg)
Introduction
Evolve algorithms with Genetic Programming.
Algorithms ≡ instructions + control flow + memory 6= Formula
Exact integer computations: Result is either right or wrong
Hard for GP
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 3/27
![Page 11: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/11.jpg)
Introduction
Evolve algorithms with Genetic Programming.
Algorithms ≡ instructions + control flow + memory 6= Formula
Exact integer computations: Result is either right or wrong
Hard for GP
rugged fitness landscape
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 3/27
![Page 12: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/12.jpg)
Introduction
Evolve algorithms with Genetic Programming.
Algorithms ≡ instructions + control flow + memory 6= Formula
Exact integer computations: Result is either right or wrong
Hard for GP
rugged fitness landscape
epistasis, deceptiveness
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 3/27
![Page 13: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/13.jpg)
Introduction
Evolve algorithms with Genetic Programming.
Algorithms ≡ instructions + control flow + memory 6= Formula
Exact integer computations: Result is either right or wrong
Hard for GP
rugged fitness landscape
epistasis, deceptiveness
Questions
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 3/27
![Page 14: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/14.jpg)
Introduction
Evolve algorithms with Genetic Programming.
Algorithms ≡ instructions + control flow + memory 6= Formula
Exact integer computations: Result is either right or wrong
Hard for GP
rugged fitness landscape
epistasis, deceptiveness
Questions:
Can this actually be done with GP?
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 3/27
![Page 15: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/15.jpg)
Introduction
Evolve algorithms with Genetic Programming.
Algorithms ≡ instructions + control flow + memory 6= Formula
Exact integer computations: Result is either right or wrong
Hard for GP
rugged fitness landscape
epistasis, deceptiveness
Questions:
Can this actually be done with GP for non-trivial problems?
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 3/27
![Page 16: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/16.jpg)
Introduction
Evolve algorithms with Genetic Programming.
Algorithms ≡ instructions + control flow + memory 6= Formula
Exact integer computations: Result is either right or wrong
Hard for GP
rugged fitness landscape
epistasis, deceptiveness
Questions:
Can this actually be done with GP for non-trivial problems?
How can we improve the chance to success?
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 3/27
![Page 17: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/17.jpg)
Introduction
Evolve algorithms with Genetic Programming.
Algorithms ≡ instructions + control flow + memory 6= Formula
Exact integer computations: Result is either right or wrong
Hard for GP
rugged fitness landscape
epistasis, deceptiveness
Questions:
Can this actually be done with GP for non-trivial problems?
How can we improve the chance to success?
Contributions
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 3/27
![Page 18: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/18.jpg)
Introduction
Evolve algorithms with Genetic Programming.
Algorithms ≡ instructions + control flow + memory 6= Formula
Exact integer computations: Result is either right or wrong
Hard for GP
rugged fitness landscape
epistasis, deceptiveness
Questions:
Can this actually be done with GP for non-trivial problems?
How can we improve the chance to success?
Contributions:
Benchmark cases
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 3/27
![Page 19: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/19.jpg)
Introduction
Evolve algorithms with Genetic Programming.
Algorithms ≡ instructions + control flow + memory 6= Formula
Exact integer computations: Result is either right or wrong
Hard for GP
rugged fitness landscape
epistasis, deceptiveness
Questions:
Can this actually be done with GP for non-trivial problems?
How can we improve the chance to success?
Contributions:
Benchmark cases
2 technologies, 1 works, 1 doesn’t
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 3/27
![Page 20: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/20.jpg)
1 Benchmark Problems
2 Investigated Ideas
3 Experiments
4 Summary
5 Evolved Programs
Benchmark ProblemsEvolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 4/27
![Page 21: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/21.jpg)
Benchmark Problems
A benchmark problem in GP is defined by
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 5/27
![Page 22: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/22.jpg)
Benchmark Problems
A benchmark problem in GP is defined by:
1 the instruction set,
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 5/27
![Page 23: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/23.jpg)
Benchmark Problems
A benchmark problem in GP is defined by:
1 the instruction set,
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 5/27
![Page 24: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/24.jpg)
Benchmark Problems
A benchmark problem in GP is defined by:
1 the instruction set,2 the objective function f ,
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 5/27
![Page 25: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/25.jpg)
Benchmark Problems
A benchmark problem in GP is defined by:
1 the instruction set,2 the objective function f ,3 the algorithm ϕ that we want to evolve,
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 5/27
![Page 26: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/26.jpg)
Benchmark Problems
A benchmark problem in GP is defined by:
1 the instruction set,2 the objective function f ,3 the algorithm ϕ that we want to evolve (and tc corresponding training
cases ti),
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 5/27
![Page 27: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/27.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set:
![Page 28: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/28.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +
![Page 29: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/29.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -
![Page 30: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/30.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -, *
![Page 31: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/31.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -, *, /
![Page 32: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/32.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -, *, /, %
![Page 33: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/33.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -, *, /, %, E ∈ {0, 1}
![Page 34: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/34.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -, *, /, %, E ∈ {0, 1}, m1 m2 m3 m4 m5
![Page 35: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/35.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -, *, /, %, E ∈ {0, 1}, m1 m2 m3 m4 m5, =
![Page 36: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/36.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -, *, /, %, E ∈ {0, 1}, m1 m2 m3 m4 m5, =, ;
![Page 37: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/37.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -, *, /, %, E ∈ {0, 1}, m1 m2 m3 m4 m5, =, ;, loop
![Page 38: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/38.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -, *, /, %, E ∈ {0, 1}, m1 m2 m3 m4 m5, =, ;, loop
Objective Function: f(x) = Error Rate
![Page 39: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/39.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -, *, /, %, E ∈ {0, 1}, m1 m2 m3 m4 m5, =, ;, loop
Objective Function: f(x) = # training cases ti with ϕ(ti) 6= x(ti)
![Page 40: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/40.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -, *, /, %, E ∈ {0, 1}, m1 m2 m3 m4 m5, =, ;, loop
Objective Function: f(x) = # training cases ti with ϕ(ti) 6= x(ti)total number tc of training cases
![Page 41: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/41.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -, *, ∅/, %, E ∈ {0, 1}, m1 m2∅m3∅m4∅m5, =, ;, loop
Objective Function: f(x) = # training cases ti with ϕ(ti) 6= x(ti)total number tc=100 of training cases
1 po2 [1]: ϕ1(ti) = t3i + t2i + 2 ∗ ti
![Page 42: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/42.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -, *, ∅/, %, E ∈ {0, 1}, m1 m2∅m3∅m4∅m5, =, ;, loop
Objective Function: f(x) = # training cases ti with ϕ(ti) 6= x(ti)total number tc=100 of training cases
1 po2 [1]: ϕ1(ti) = t3i + t2i + 2 ∗ ti2 su2
[1]: sum of first ti natural numbers
![Page 43: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/43.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -, *, ∅/, %, E ∈ {0, 1}, m1 m2∅m3∅m4∅m5, =, ;, loop
Objective Function: f(x) = # training cases ti with ϕ(ti) 6= x(ti)total number tc=12 of training cases
1 po2 [1]: ϕ1(ti) = t3i + t2i + 2 ∗ ti2 su2
[1]: sum of first ti natural numbers
3 fac [1, 2]: ϕ3(ti) = ti!
![Page 44: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/44.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -, *, ∅/, %, E ∈ {0, 1}, m1 m2 m3∅m4∅m5, =, ;, loop
Objective Function: f(x) = # training cases ti with ϕ(ti) 6= x(ti)total number tc=100 of training cases
1 po2 [1]: ϕ1(ti) = t3i + t2i + 2 ∗ ti2 su2
[1]: sum of first ti natural numbers
3 fac [1, 2]: ϕ3(ti) = ti!
4 po3, su3: po2, su2 with 3 memory cells
![Page 45: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/45.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -, *, ∅/, %, E ∈ {0, 1}, m1 m2 m3∅m4∅m5, =, ;, loop
Objective Function: f(x) = # training cases ti with ϕ(ti) 6= x(ti)total number tc=100 of training cases
1 po2 [1]: ϕ1(ti) = t3i + t2i + 2 ∗ ti2 su2
[1]: sum of first ti natural numbers
3 fac [1, 2]: ϕ3(ti) = ti!
4 po3, su3: po2, su2 with 3 memory cells
6 gcd [1, 3]: compute greatest common divisor
![Page 46: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/46.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -, *, /, %, E ∈ {0, 1}, m1 m2 m3∅m4∅m5, =, ;, loop
Objective Function: f(x) = # training cases ti with ϕ(ti) 6= x(ti)total number tc=31 of training cases
1 po2 [1]: ϕ1(ti) = t3i + t2i + 2 ∗ ti2 su2
[1]: sum of first ti natural numbers
3 fac [1, 2]: ϕ3(ti) = ti!
4 po3, su3: po2, su2 with 3 memory cells
6 gcd [1, 3]: compute greatest common divisor
7 exp: ϕ7(ti) = 2ti
![Page 47: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/47.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -, *, /, %, E ∈ {0, 1}, m1 m2 m3∅m4∅m5, =, ;, loop
Objective Function: f(x) = # training cases ti with ϕ(ti) 6= x(ti)total number tc=40 of training cases
1 po2 [1]: ϕ1(ti) = t3i + t2i + 2 ∗ ti2 su2
[1]: sum of first ti natural numbers
3 fac [1, 2]: ϕ3(ti) = ti!
4 po3, su3: po2, su2 with 3 memory cells
6 gcd [1, 3]: compute greatest common divisor
7 exp: ϕ7(ti) = 2ti
8 ℓ20: return 0 if ti < 20, 1 otherwise
![Page 48: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/48.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -, *, /, %, E ∈ {0, 1}, m1 m2 m3∅m4∅m5, =, ;, loop
Objective Function: f(x) = # training cases ti with ϕ(ti) 6= x(ti)total number tc=100 of training cases
1 po2 [1]: ϕ1(ti) = t3i + t2i + 2 ∗ ti2 su2
[1]: sum of first ti natural numbers
3 fac [1, 2]: ϕ3(ti) = ti!
4 po3, su3: po2, su2 with 3 memory cells
6 gcd [1, 3]: compute greatest common divisor
7 exp: ϕ7(ti) = 2ti
8 ℓ20: return 0 if ti < 20, 1 otherwise
9 prm: returns 1 if ti is prime, 0
![Page 49: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/49.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -, *, /, %, E ∈ {0, 1}, m1 m2 m3∅m4∅m5, =, ;, loop
Objective Function: f(x) = # training cases ti with ϕ(ti) 6= x(ti)total number tc=100 of training cases
1 po2 [1]: ϕ1(ti) = t3i + t2i + 2 ∗ ti2 su2
[1]: sum of first ti natural numbers
3 fac [1, 2]: ϕ3(ti) = ti!
4 po3, su3: po2, su2 with 3 memory cells
6 gcd [1, 3]: compute greatest common divisor
7 exp: ϕ7(ti) = 2ti
8 ℓ20: return 0 if ti < 20, 1 otherwise
9 prm: returns 1 if ti is prime, 0
10 ssq: ϕ10(ti) =∑ti
i=1i2
![Page 50: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/50.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -, *, /, %, E ∈ {0, 1}, m1 m2 m3∅m4∅m5, =, ;, loop
Objective Function: f(x) = # training cases ti with ϕ(ti) 6= x(ti)total number tc=100 of training cases
1 po2 [1]: ϕ1(ti) = t3i + t2i + 2 ∗ ti2 su2
[1]: sum of first ti natural numbers
3 fac [1, 2]: ϕ3(ti) = ti!
4 po3, su3: po2, su2 with 3 memory cells
6 gcd [1, 3]: compute greatest common divisor
7 exp: ϕ7(ti) = 2ti
8 ℓ20: return 0 if ti < 20, 1 otherwise
9 prm: returns 1 if ti is prime, 0
10 ssq: ϕ10(ti) =∑ti
i=1i2
11 sra: ϕ11(ti) =√ti
![Page 51: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/51.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -, *, /, %, E ∈ {0, 1}, m1 m2 m3∅m4∅m5, =, ;, loop
Objective Function: f(x) = # training cases ti with ϕ(ti) 6= x(ti)total number tc=100 of training cases
1 po2 [1]: ϕ1(ti) = t3i + t2i + 2 ∗ ti2 su2
[1]: sum of first ti natural numbers
3 fac [1, 2]: ϕ3(ti) = ti!
4 po3, su3: po2, su2 with 3 memory cells
6 gcd [1, 3]: compute greatest common divisor
7 exp: ϕ7(ti) = 2ti
8 ℓ20: return 0 if ti < 20, 1 otherwise
9 prm: returns 1 if ti is prime, 0
10 ssq: ϕ10(ti) =∑ti
i=1i2
11 sra: ϕ11(ti) =√ti
12 srb: ϕ12(ti) =⌊√
ti⌋
![Page 52: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/52.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -, *, /, %, E ∈ {0, 1}, m1 m2 m3∅m4∅m5, =, ;, loop
Objective Function: f(x) = # training cases ti with ϕ(ti) 6= x(ti)total number tc=100 of training cases
1 po2 [1]: ϕ1(ti) = t3i + t2i + 2 ∗ ti2 su2
[1]: sum of first ti natural numbers
3 fac [1, 2]: ϕ3(ti) = ti!
4 po3, su3: po2, su2 with 3 memory cells
6 gcd [1, 3]: compute greatest common divisor
7 exp: ϕ7(ti) = 2ti
8 ℓ20: return 0 if ti < 20, 1 otherwise
9 prm: returns 1 if ti is prime, 0
10 ssq: ϕ10(ti) =∑ti
i=1i2
11 sra: ϕ11(ti) =√ti
12 srb: ϕ12(ti) =⌊√
ti⌋
13 ild: ld⋆(y) ={
0 if ti ≤ 11 + ld⋆(ld(y)) otherwise
![Page 53: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/53.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -, *, /, %, E ∈ {0, 1}, m1 m2 m3∅m4∅m5, =, ;, loop
Objective Function: f(x) = # training cases ti with ϕ(ti) 6= x(ti)total number tc=100 of training cases
1 po2 [1]: ϕ1(ti) = t3i + t2i + 2 ∗ ti2 su2
[1]: sum of first ti natural numbers
3 fac [1, 2]: ϕ3(ti) = ti!
4 po3, su3: po2, su2 with 3 memory cells
6 gcd [1, 3]: compute greatest common divisor
7 exp: ϕ7(ti) = 2ti
8 ℓ20: return 0 if ti < 20, 1 otherwise
9 prm: returns 1 if ti is prime, 0
10 ssq: ϕ10(ti) =∑ti
i=1i2
11 sra: ϕ11(ti) =√ti
12 srb: ϕ12(ti) =⌊√
ti⌋
13 ild: ld⋆(y) ={
0 if ti ≤ 11 + ld⋆(ld(y)) otherwise
14 lsb: least significant bit in the two’s
complement
![Page 54: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/54.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -, ∅*, ∅/, ∅%, E ∈ {0, 1}, m1 m2 m3∅m4∅m5, =, ;, loop
Objective Function: f(x) = # training cases ti with ϕ(ti) 6= x(ti)total number tc=100 of training cases
1 po2 [1]: ϕ1(ti) = t3i + t2i + 2 ∗ ti2 su2
[1]: sum of first ti natural numbers
3 fac [1, 2]: ϕ3(ti) = ti!
4 po3, su3: po2, su2 with 3 memory cells
6 gcd [1, 3]: compute greatest common divisor
7 exp: ϕ7(ti) = 2ti
8 ℓ20: return 0 if ti < 20, 1 otherwise
9 prm: returns 1 if ti is prime, 0
10 ssq: ϕ10(ti) =∑ti
i=1i2
11 sra: ϕ11(ti) =√ti
12 srb: ϕ12(ti) =⌊√
ti⌋
13 ild: ld⋆(y) ={
0 if ti ≤ 11 + ld⋆(ld(y)) otherwise
14 lsb: least significant bit in the two’s
complement
15 mul: ϕ15(ti) = ti,1 ∗ ti,2
![Page 55: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/55.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -, ∅*, ∅/, ∅%, E ∈ {0, 1}, m1 m2 m3∅m4∅m5, =, ;, loop
Objective Function: f(x) = # training cases ti with ϕ(ti) 6= x(ti)total number tc=100 of training cases
1 po2 [1]: ϕ1(ti) = t3i + t2i + 2 ∗ ti2 su2
[1]: sum of first ti natural numbers
3 fac [1, 2]: ϕ3(ti) = ti!
4 po3, su3: po2, su2 with 3 memory cells
6 gcd [1, 3]: compute greatest common divisor
7 exp: ϕ7(ti) = 2ti
8 ℓ20: return 0 if ti < 20, 1 otherwise
9 prm: returns 1 if ti is prime, 0
10 ssq: ϕ10(ti) =∑ti
i=1i2
11 sra: ϕ11(ti) =√ti
12 srb: ϕ12(ti) =⌊√
ti⌋
13 ild: ld⋆(y) ={
0 if ti ≤ 11 + ld⋆(ld(y)) otherwise
14 lsb: least significant bit in the two’s
complement
15 mul: ϕ15(ti) = ti,1 ∗ ti,216 qad: ϕ16(ti) = (ti − 1)(ti + 2)
![Page 56: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/56.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -, ∅*, ∅/, ∅%, E ∈ {0, 1}, m1 m2 m3∅m4∅m5, =, ;, loop
Objective Function: f(x) = # training cases ti with ϕ(ti) 6= x(ti)total number tc=100 of training cases
1 po2 [1]: ϕ1(ti) = t3i + t2i + 2 ∗ ti2 su2
[1]: sum of first ti natural numbers
3 fac [1, 2]: ϕ3(ti) = ti!
4 po3, su3: po2, su2 with 3 memory cells
6 gcd [1, 3]: compute greatest common divisor
7 exp: ϕ7(ti) = 2ti
8 ℓ20: return 0 if ti < 20, 1 otherwise
9 prm: returns 1 if ti is prime, 0
10 ssq: ϕ10(ti) =∑ti
i=1i2
11 sra: ϕ11(ti) =√ti
12 srb: ϕ12(ti) =⌊√
ti⌋
13 ild: ld⋆(y) ={
0 if ti ≤ 11 + ld⋆(ld(y)) otherwise
14 lsb: least significant bit in the two’s
complement
15 mul: ϕ15(ti) = ti,1 ∗ ti,216 qad: ϕ16(ti) = (ti − 1)(ti + 2)
17 mod: ϕ17(ti) = ti,1 mod ti,2
![Page 57: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/57.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -, *, /, %, E ∈ {0, 1}, m1 m2 m3 m4 m5, =, ;, loop
Objective Function: f(x) = # training cases ti with ϕ(ti) 6= x(ti)total number tc=100 of training cases
1 po2 [1]: ϕ1(ti) = t3i + t2i + 2 ∗ ti2 su2
[1]: sum of first ti natural numbers
3 fac [1, 2]: ϕ3(ti) = ti!
4 po3, su3: po2, su2 with 3 memory cells
6 gcd [1, 3]: compute greatest common divisor
7 exp: ϕ7(ti) = 2ti
8 ℓ20: return 0 if ti < 20, 1 otherwise
9 prm: returns 1 if ti is prime, 0
10 ssq: ϕ10(ti) =∑ti
i=1i2
11 sra: ϕ11(ti) =√ti
12 srb: ϕ12(ti) =⌊√
ti⌋
13 ild: ld⋆(y) ={
0 if ti ≤ 11 + ld⋆(ld(y)) otherwise
14 lsb: least significant bit in the two’s
complement
15 mul: ϕ15(ti) = ti,1 ∗ ti,216 qad: ϕ16(ti) = (ti − 1)(ti + 2)
17 mod: ϕ17(ti) = ti,1 mod ti,2
18 mi5: minimum of five
![Page 58: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/58.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -, *, /, %, E ∈ {0, 1}, m1 m2 m3 m4 m5, =, ;, loop
Objective Function: f(x) = # training cases ti with ϕ(ti) 6= x(ti)total number tc=100 of training cases
1 po2 [1]: ϕ1(ti) = t3i + t2i + 2 ∗ ti2 su2
[1]: sum of first ti natural numbers
3 fac [1, 2]: ϕ3(ti) = ti!
4 po3, su3: po2, su2 with 3 memory cells
6 gcd [1, 3]: compute greatest common divisor
7 exp: ϕ7(ti) = 2ti
8 ℓ20: return 0 if ti < 20, 1 otherwise
9 prm: returns 1 if ti is prime, 0
10 ssq: ϕ10(ti) =∑ti
i=1i2
11 sra: ϕ11(ti) =√ti
12 srb: ϕ12(ti) =⌊√
ti⌋
13 ild: ld⋆(y) ={
0 if ti ≤ 11 + ld⋆(ld(y)) otherwise
14 lsb: least significant bit in the two’s
complement
15 mul: ϕ15(ti) = ti,1 ∗ ti,216 qad: ϕ16(ti) = (ti − 1)(ti + 2)
17 mod: ϕ17(ti) = ti,1 mod ti,2
18 mi5: minimum of five
19 ma5: maximum of five
![Page 59: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/59.jpg)
Benchmark Problems
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 6/27
Instruction Set: +, -, *, /, %, E ∈ {0, 1}, m1 m2 m3 m4 m5, =, ;, loop
Objective Function: f(x) = # training cases ti with ϕ(ti) 6= x(ti)total number tc=100 of training cases
1 po2 [1]: ϕ1(ti) = t3i + t2i + 2 ∗ ti2 su2
[1]: sum of first ti natural numbers
3 fac [1, 2]: ϕ3(ti) = ti!
4 po3, su3: po2, su2 with 3 memory cells
6 gcd [1, 3]: compute greatest common divisor
7 exp: ϕ7(ti) = 2ti
8 ℓ20: return 0 if ti < 20, 1 otherwise
9 prm: returns 1 if ti is prime, 0
10 ssq: ϕ10(ti) =∑ti
i=1i2
11 sra: ϕ11(ti) =√ti
12 srb: ϕ12(ti) =⌊√
ti⌋
13 ild: ld⋆(y) ={
0 if ti ≤ 11 + ld⋆(ld(y)) otherwise
14 lsb: least significant bit in the two’s
complement
15 mul: ϕ15(ti) = ti,1 ∗ ti,216 qad: ϕ16(ti) = (ti − 1)(ti + 2)
17 mod: ϕ17(ti) = ti,1 mod ti,2
18 mi5: minimum of five
19 ma5: maximum of five
20 sm5: sum of five
![Page 60: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/60.jpg)
1 Benchmark Problems
2 Investigated Ideas
3 Experiments
4 Summary
5 Evolved Programs
Investigated IdeasEvolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 7/27
![Page 61: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/61.jpg)
Loop Instructions
We test six different loop instructions
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 8/27
![Page 62: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/62.jpg)
Loop Instructions
We test six different loop instructions
1 CL: execute loop body n times
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 8/27
![Page 63: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/63.jpg)
Loop Instructions
We test six different loop instructions
1 CL: execute loop body n times
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 8/27
![Page 64: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/64.jpg)
Loop Instructions
We test six different loop instructions
1 CL: execute loop body n times2 ML: decrease memory cell in each iteration,
stop when 0 is reached
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 8/27
![Page 65: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/65.jpg)
Loop Instructions
We test six different loop instructions
1 CL: execute loop body n times2 ML: decrease memory cell in each iteration,
stop when 0 is reached3 WL: execute until expression becomes 0
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 8/27
![Page 66: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/66.jpg)
Loop Instructions
We test six different loop instructions
1 CL: execute loop body n times2 ML: decrease memory cell in each iteration,
stop when 0 is reached3 WL: execute until expression becomes 04 CA: conditional “=” assignment, execute
program as long as variables change
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 8/27
![Page 67: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/67.jpg)
Loop Instructions
We test six different loop instructions
1 CL: execute loop body n times2 ML: decrease memory cell in each iteration,
stop when 0 is reached3 WL: execute until expression becomes 04 CA: conditional “=” assignment, execute
program as long as variables change5 IL: execute loop body (and program) as long
as variables change
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 8/27
![Page 68: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/68.jpg)
Loop Instructions
We test six different loop instructions
1 CL: execute loop body n times2 ML: decrease memory cell in each iteration,
stop when 0 is reached3 WL: execute until expression becomes 04 CA: conditional “=” assignment, execute
program as long as variables change5 IL: execute loop body (and program) as long
as variables change6 IC: IL ⊕ CA
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 8/27
![Page 69: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/69.jpg)
Fighting Epistasis: Transactional Memory
Program behavior depends on
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 9/27
![Page 70: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/70.jpg)
Fighting Epistasis: Transactional Memory
Program behavior depends on instructions
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 9/27
![Page 71: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/71.jpg)
Fighting Epistasis: Transactional Memory
Program behavior depends on a) instructions and b) their order
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 9/27
![Page 72: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/72.jpg)
Fighting Epistasis: Transactional Memory
Program behavior depends on a) instructions and b) their order
=⇒ epistasis!
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 9/27
![Page 73: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/73.jpg)
Fighting Epistasis: Transactional Memory
Program behavior depends on a) instructions and b) their order
=⇒ epistasis!
Transactional Memory (TM) [4]
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 9/27
![Page 74: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/74.jpg)
Fighting Epistasis: Transactional Memory
Program behavior depends on a) instructions and b) their order
=⇒ epistasis!
Transactional Memory (TM) [4]
hierarchical TM: each control scope (;, loop) has
two memory records
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 9/27
m = m + 10 1
m = m + 21 0
![Page 75: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/75.jpg)
Fighting Epistasis: Transactional Memory
Program behavior depends on a) instructions and b) their order
=⇒ epistasis!
Transactional Memory (TM) [4]
hierarchical TM: each control scope (;, loop) has
two memory records
write memory: changes by write instruction =
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 9/27
m =0 m + 11
m =1 2m +0
![Page 76: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/76.jpg)
Fighting Epistasis: Transactional Memory
Program behavior depends on a) instructions and b) their order
=⇒ epistasis!
Transactional Memory (TM) [4]
hierarchical TM: each control scope (;, loop) has
two memory records
write memory: changes by write instruction =
read memory: values returned by variable ac-
cess mi
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 9/27
m =0 + 1m1
m =1 + 2m0
![Page 77: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/77.jpg)
Fighting Epistasis: Transactional Memory
Program behavior depends on a) instructions and b) their order
=⇒ epistasis!
Transactional Memory (TM) [4]
hierarchical TM: each control scope (;, loop) has
two memory records
write memory: changes by write instruction =
read memory: values returned by variable ac-
cess mi
write memory is comitted at end of control scope
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 9/27
m =0 + 1m1
m =1 + 2m0
(commit)
![Page 78: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/78.jpg)
Fighting Epistasis: Transactional Memory
Program behavior depends on a) instructions and b) their order
=⇒ epistasis!
Transactional Memory (TM) [4]
hierarchical TM: each control scope (;, loop) has
two memory records
write memory: changes by write instruction =
read memory: values returned by variable ac-
cess mi
write memory is comitted to memory of parent
scope at end of control scope
Order of instructions becomes less important
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 9/27
m =0 + 1m1
m =1 + 2m0
(commit)
![Page 79: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/79.jpg)
Fighting Epistasis: Transactional Memory
Program behavior depends on a) instructions and b) their order
=⇒ epistasis!
Transactional Memory (TM) [4]
hierarchical TM: each control scope (;, loop) has
two memory records
write memory: changes by write instruction =
read memory: values returned by variable ac-
cess mi
write memory is comitted to memory of parent
scope at end of control scope
Order of instructions becomes less important
Can this reduce epistasis?
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 9/27
m =0 + 1m1
m =1 + 2m0
(commit)
m =0 + 1m1
m =1 + 2m0
(commit)
![Page 80: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/80.jpg)
Fighting Epistasis: Transactional Memory
Program behavior depends on a) instructions and b) their order
=⇒ epistasis!
Transactional Memory (TM) [4]
hierarchical TM: each control scope (;, loop) has
two memory records
write memory: changes by write instruction =
read memory: values returned by variable ac-
cess mi
write memory is comitted to memory of parent
scope at end of control scope
Order of instructions becomes less important
Can this reduce epistasis?
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 9/27
m =0 + 1m1
m =1 + 2m0
(commit)
m =0 + 1m1
m =1 + 2m0
(commit)
![Page 81: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/81.jpg)
Fighting Deceptiveness: FFA
Genetic Programming minimizes f
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 10/27
![Page 82: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/82.jpg)
Fighting Deceptiveness: FFA
Genetic Programming minimizes f
. . . but: Is a program x1 that can solve 60% of the training cases
closer to a solution than a program x2 that can solve 10%?
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 10/27
![Page 83: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/83.jpg)
Fighting Deceptiveness: FFA
Genetic Programming minimizes f
. . . but: Is a program x1 that can solve 60% of the training casescloser to a solution than a program x2 that can solve 10%?
Not necessarily
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 10/27
![Page 84: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/84.jpg)
Fighting Deceptiveness: FFA
Genetic Programming minimizes f
. . . but: Is a program x1 that can solve 60% of the training casescloser to a solution than a program x2 that can solve 10%?
Not necessarily: deceptiveness, many local optima
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 10/27
![Page 85: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/85.jpg)
Fighting Deceptiveness: FFA
Genetic Programming minimizes f
. . . but: Is a program x1 that can solve 60% of the training casescloser to a solution than a program x2 that can solve 10%?
Not necessarily: deceptiveness, many local optima
But selection is very likely to prefer x1!
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 10/27
![Page 86: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/86.jpg)
Fighting Deceptiveness: FFA
Genetic Programming minimizes f
. . . but: Is a program x1 that can solve 60% of the training casescloser to a solution than a program x2 that can solve 10%?
Not necessarily: deceptiveness, many local optima
But selection is very likely to prefer x1!
GP will sooner or later converge.
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 10/27
![Page 87: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/87.jpg)
Fighting Deceptiveness: FFA
Genetic Programming minimizes f
. . . but: Is a program x1 that can solve 60% of the training cases
closer to a solution than a program x2 that can solve 10%?
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 10/27
![Page 88: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/88.jpg)
Fighting Deceptiveness: FFA
Genetic Programming minimizes f
. . . but: Is a program x1 that can solve 60% of the training cases
closer to a solution than a program x2 that can solve 10%?
Prevent convergence: Frequency Fitness Assignment (FFA) [5]
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 10/27
![Page 89: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/89.jpg)
Fighting Deceptiveness: FFA
Genetic Programming minimizes f
. . . but: Is a program x1 that can solve 60% of the training cases
closer to a solution than a program x2 that can solve 10%?
Prevent convergence: Frequency Fitness Assignment (FFA) [5]
Favour new behaviors instead of “good” ones
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 10/27
![Page 90: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/90.jpg)
Fighting Deceptiveness: FFA
Genetic Programming minimizes f
. . . but: Is a program x1 that can solve 60% of the training cases
closer to a solution than a program x2 that can solve 10%?
Prevent convergence: Frequency Fitness Assignment (FFA) [5]
Favour new behaviors instead of “good” ones
f defines behavioral semantics (to some degree)
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 10/27
![Page 91: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/91.jpg)
Fighting Deceptiveness: FFA
Genetic Programming minimizes f
. . . but: Is a program x1 that can solve 60% of the training cases
closer to a solution than a program x2 that can solve 10%?
Prevent convergence: Frequency Fitness Assignment (FFA) [5]
Favour new behaviors instead of “good” ones
f defines behavioral semantics (to some degree)
tc training cases =⇒ tc+ 1 possible f values
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 10/27
![Page 92: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/92.jpg)
Fighting Deceptiveness: FFA
Genetic Programming minimizes f
. . . but: Is a program x1 that can solve 60% of the training cases
closer to a solution than a program x2 that can solve 10%?
Prevent convergence: Frequency Fitness Assignment (FFA) [5]
Favour new behaviors instead of “good” ones
f defines behavioral semantics (to some degree)
tc training cases =⇒ tc+ 1 possible behaviors
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 10/27
![Page 93: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/93.jpg)
Fighting Deceptiveness: FFA
FFA-Genetic Programming minimizes H
. . . but: Is a program x1 that can solve 60% of the training cases
closer to a solution than a program x2 that can solve 10%?
Prevent convergence: Frequency Fitness Assignment (FFA) [5]
Favour new behaviors instead of “good” ones
f defines behavioral semantics (to some degree)
tc training cases =⇒ tc+ 1 possible behaviors
Fitness H [f(x)]: How often has a given f value been encountered
during the GP process?
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 10/27
![Page 94: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/94.jpg)
Fighting Deceptiveness: FFA
FFA-Genetic Programming minimizes H
. . . but: Is a program x1 that can solve 60% of the training cases
closer to a solution than a program x2 that can solve 10%?
Prevent convergence: Frequency Fitness Assignment (FFA) [5]
Favour new behaviors instead of “good” ones
f defines behavioral semantics (to some degree)
tc training cases =⇒ tc+ 1 possible behaviors
Fitness H [f(x)]: How often has a given f value been encountered
during the GP process?
high H-value: behavior seen often
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 10/27
![Page 95: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/95.jpg)
Fighting Deceptiveness: FFA
FFA-Genetic Programming minimizes H
. . . but: Is a program x1 that can solve 60% of the training cases
closer to a solution than a program x2 that can solve 10%?
Prevent convergence: Frequency Fitness Assignment (FFA) [5]
Favour new behaviors instead of “good” ones
f defines behavioral semantics (to some degree)
tc training cases =⇒ tc+ 1 possible behaviors
Fitness H [f(x)]: How often has a given f value been encountered
during the GP process?
high H-value: behavior seen often =⇒ not new
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 10/27
![Page 96: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/96.jpg)
Fighting Deceptiveness: FFA
FFA-Genetic Programming minimizes H
. . . but: Is a program x1 that can solve 60% of the training cases
closer to a solution than a program x2 that can solve 10%?
Prevent convergence: Frequency Fitness Assignment (FFA) [5]
Favour new behaviors instead of “good” ones
f defines behavioral semantics (to some degree)
tc training cases =⇒ tc+ 1 possible behaviors
Fitness H [f(x)]: How often has a given f value been encountered
during the GP process?
high H-value: behavior seen often =⇒ not new =⇒ simple to discover
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 10/27
![Page 97: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/97.jpg)
Fighting Deceptiveness: FFA
FFA-Genetic Programming minimizes H
. . . but: Is a program x1 that can solve 60% of the training cases
closer to a solution than a program x2 that can solve 10%?
Prevent convergence: Frequency Fitness Assignment (FFA) [5]
Favour new behaviors instead of “good” ones
f defines behavioral semantics (to some degree)
tc training cases =⇒ tc+ 1 possible behaviors
Fitness H [f(x)]: How often has a given f value been encountered
during the GP process?
high H-value: behavior seen often =⇒ not new =⇒ simple to discover
=⇒ uninteresting
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 10/27
![Page 98: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/98.jpg)
Fighting Deceptiveness: FFA
FFA-Genetic Programming minimizes H
. . . but: Is a program x1 that can solve 60% of the training cases
closer to a solution than a program x2 that can solve 10%?
Prevent convergence: Frequency Fitness Assignment (FFA) [5]
Favour new behaviors instead of “good” ones
f defines behavioral semantics (to some degree)
tc training cases =⇒ tc+ 1 possible behaviors
Fitness H [f(x)]: How often has a given f value been encountered
during the GP process?
high H-value: behavior seen often =⇒ not new =⇒ simple to discover
=⇒ uninteresting
low H-value: behavior new
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 10/27
![Page 99: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/99.jpg)
Fighting Deceptiveness: FFA
FFA-Genetic Programming minimizes H
. . . but: Is a program x1 that can solve 60% of the training cases
closer to a solution than a program x2 that can solve 10%?
Prevent convergence: Frequency Fitness Assignment (FFA) [5]
Favour new behaviors instead of “good” ones
f defines behavioral semantics (to some degree)
tc training cases =⇒ tc+ 1 possible behaviors
Fitness H [f(x)]: How often has a given f value been encountered
during the GP process?
high H-value: behavior seen often =⇒ not new =⇒ simple to discover
=⇒ uninteresting
low H-value: behavior new =⇒ program does something which may
be worth exploring
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 10/27
![Page 100: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/100.jpg)
Fighting Deceptiveness: FFA
FFA-Genetic Programming minimizes H
. . . but: Is a program x1 that can solve 60% of the training cases
closer to a solution than a program x2 that can solve 10%?
Prevent convergence: Frequency Fitness Assignment (FFA) [5]
Favour new behaviors instead of “good” ones
f defines behavioral semantics (to some degree)
tc training cases =⇒ tc+ 1 possible behaviors
Fitness H [f(x)]: How often has a given f value been encountered
during the GP process?
high H-value: behavior seen often =⇒ not new =⇒ simple to discover
=⇒ uninteresting
low H-value: behavior new =⇒ program does something which may
be worth exploring
H is a dynamic fitness function: behaviors get “worse” over time
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 10/27
![Page 101: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/101.jpg)
Fighting Deceptiveness: FFA
FFA-Genetic Programming minimizes H
. . . but: Is a program x1 that can solve 60% of the training cases
closer to a solution than a program x2 that can solve 10%?
Prevent convergence: Frequency Fitness Assignment (FFA) [5]
Can FFA improve exploration and fight deceptiveness?
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 10/27
![Page 102: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/102.jpg)
1 Benchmark Problems
2 Investigated Ideas
3 Experiments
4 Summary
5 Evolved Programs
ExperimentsEvolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 11/27
![Page 103: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/103.jpg)
Experiments
On each of the new benchmarks, we tested
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 12/27
![Page 104: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/104.jpg)
Experiments
On each of the new benchmarks, we tested 6 setups
6 loop structures
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 12/27
![Page 105: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/105.jpg)
Experiments
On each of the new benchmarks, we tested 6 × 2 setups
6 loop structures
2 memory models
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 12/27
![Page 106: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/106.jpg)
Experiments
On each of the new benchmarks, we tested 6 × 2 × 2 = 24 setups
6 loop structures
2 memory models
2 fitness functions
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 12/27
![Page 107: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/107.jpg)
Experiments
On each of the new benchmarks, we tested 24 setups
6 loop structures
2 memory models
2 fitness functions for
100 runs each
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 12/27
![Page 108: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/108.jpg)
Experiments
On each of the new benchmarks, we tested 24 setups
6 loop structures
2 memory models
2 fitness functions for
100 runs each
population size 1000
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 12/27
![Page 109: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/109.jpg)
Experiments
On each of the new benchmarks, we tested 24 setups
6 loop structures
2 memory models
2 fitness functions for
100 runs each
population size 1000
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 12/27
![Page 110: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/110.jpg)
Experiments
On each of the new benchmarks, we tested 24 setups
6 loop structures
2 memory models
2 fitness functions for
100 runs each
population size 1000, 100 generations
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 12/27
![Page 111: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/111.jpg)
Experiments
On each of the new benchmarks, we tested 24 setups
6 loop structures
2 memory models
2 fitness functions for
100 runs each
population size 1000, 100 generations, tournament selection with 7
contestants
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 12/27
![Page 112: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/112.jpg)
Experiments
On each of the new benchmarks, we tested 24 setups
6 loop structures
2 memory models
2 fitness functions for
100 runs each
population size 1000, 100 generations, tournament selection with 7
contestants, 10% mutation
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 12/27
![Page 113: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/113.jpg)
Experiments
On each of the new benchmarks, we tested 24 setups
6 loop structures
2 memory models
2 fitness functions for
100 runs each
population size 1000, 100 generations, tournament selection with 7
contestants, 10% mutation, 90% crossover
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 12/27
![Page 114: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/114.jpg)
Results
48 000 runs
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 13/27
![Page 115: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/115.jpg)
Results
48 000 runs by only 5422 (11%) were successful.
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 13/27
![Page 116: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/116.jpg)
Results
48 000 runs by only 5422 (11%) were successful.
What factors contribute to success (or failure)?
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 13/27
![Page 117: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/117.jpg)
Results
48 000 runs by only 5422 (11%) were successful.
What factors contribute to success (or failure)?
Let’s look at the success rates. . .
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 13/27
![Page 118: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/118.jpg)
Setup po2 su2 po3 su3 fac gcd exp ℓ20 prm ssq sra srb ild lsb mul qad mod mi5 ma5 sm5 All
DIR 0.10 0.10 0.05 0.05 0.07 0.29 0.03 0.66 0.00 0.02 0.06 0.00 0.03 0.01 0.12 0.01 0.03 0.10 0.13 0.06 0.10
FFA 0.18 0.11 0.09 0.07 0.13 0.41 0.03 0.86 0.01 0.01 0.08 0.02 0.02 0.01 0.17 0.09 0.05 0.20 0.24 0.07 0.14
SM 0.16 0.12 0.09 0.07 0.07 0.48 0.05 0.87 0.01 0.02 0.08 0.02 0.03 0.01 0.23 0.07 0.07 0.20 0.23 0.04 0.15
TM 0.12 0.09 0.05 0.05 0.13 0.22 0.00 0.65 0.00 0.01 0.06 0.00 0.02 0.01 0.06 0.04 0.01 0.09 0.14 0.09 0.09
SM-DIR-CL 0.14 0.03 0.04 0.01 0.01 0.37 0.20 0.93 0.00 0.00 0.01 0.00 0.00 0.00 0.44 0.07 0.22 0.04 0.07 0.11 0.13
SM-DIR-ML 0.18 0.66 0.08 0.32 0.27 0.36 0.11 0.70 0.00 0.21 0.03 0.00 0.01 0.00 0.63 0.01 0.13 0.00 0.00 0.04 0.19
SM-DIR-WL 0.11 0.00 0.08 0.00 0.00 0.42 0.00 0.51 0.00 0.00 0.01 0.00 0.00 0.01 0.00 0.00 0.00 0.00 0.00 0.07 0.06
SM-DIR-CA 0.11 0.00 0.08 0.00 0.00 0.55 0.00 0.93 0.01 0.00 0.14 0.01 0.13 0.03 0.00 0.00 0.00 0.40 0.50 0.00 0.14
SM-DIR-IL 0.09 0.01 0.11 0.01 0.00 0.10 0.00 0.81 0.00 0.00 0.04 0.00 0.01 0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.06
SM-DIR-IC 0.04 0.03 0.00 0.00 0.00 0.68 0.00 0.88 0.01 0.00 0.13 0.02 0.05 0.02 0.01 0.00 0.00 0.43 0.55 0.02 0.14
SM-FFA-CL 0.32 0.03 0.16 0.00 0.03 0.41 0.19 1.00 0.00 0.00 0.01 0.07 0.00 0.00 0.72 0.41 0.31 0.19 0.17 0.05 0.20
SM-FFA-ML 0.14 0.57 0.15 0.46 0.45 0.47 0.12 0.92 0.02 0.05 0.05 0.00 0.00 0.00 0.89 0.30 0.15 0.00 0.00 0.06 0.24
SM-FFA-WL 0.32 0.00 0.17 0.02 0.04 0.46 0.00 0.90 0.01 0.00 0.01 0.01 0.05 0.01 0.00 0.00 0.00 0.00 0.02 0.06 0.10
SM-FFA-CA 0.17 0.00 0.05 0.00 0.02 0.80 0.00 0.97 0.02 0.00 0.22 0.06 0.02 0.01 0.00 0.00 0.00 0.64 0.74 0.00 0.19
SM-FFA-IL 0.21 0.03 0.11 0.00 0.00 0.42 0.01 0.93 0.01 0.00 0.15 0.01 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.05 0.10
SM-FFA-IC 0.12 0.02 0.01 0.03 0.00 0.76 0.00 0.90 0.05 0.00 0.16 0.03 0.01 0.01 0.01 0.00 0.00 0.72 0.74 0.04 0.18
TM-DIR-CL 0.12 0.07 0.07 0.03 0.01 0.00 0.00 0.81 0.00 0.00 0.00 0.00 0.00 0.00 0.26 0.05 0.04 0.00 0.00 0.13 0.08
TM-DIR-ML 0.17 0.27 0.04 0.14 0.43 0.01 0.00 0.30 0.00 0.01 0.18 0.00 0.03 0.00 0.03 0.00 0.01 0.00 0.00 0.14 0.09
TM-DIR-WL 0.15 0.00 0.07 0.00 0.00 0.11 0.00 0.24 0.00 0.00 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.01 0.03 0.08 0.04
TM-DIR-CA 0.04 0.05 0.04 0.08 0.05 0.83 0.00 0.66 0.00 0.00 0.11 0.01 0.06 0.01 0.04 0.01 0.00 0.24 0.30 0.10 0.13
TM-DIR-IL 0.03 0.01 0.02 0.00 0.00 0.04 0.00 0.75 0.00 0.00 0.02 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.01 0.00 0.05
TM-DIR-IC 0.04 0.03 0.00 0.01 0.09 0.05 0.00 0.45 0.00 0.00 0.04 0.00 0.03 0.02 0.00 0.00 0.00 0.05 0.15 0.06 0.05
TM-FFA-CL 0.20 0.04 0.12 0.06 0.01 0.00 0.02 1.00 0.00 0.00 0.00 0.00 0.01 0.00 0.34 0.32 0.07 0.00 0.00 0.09 0.11
TM-FFA-ML 0.24 0.28 0.09 0.13 0.58 0.00 0.00 0.72 0.00 0.03 0.05 0.00 0.01 0.00 0.07 0.09 0.04 0.00 0.00 0.12 0.12
TM-FFA-WL 0.17 0.00 0.15 0.00 0.00 0.17 0.00 0.74 0.01 0.00 0.01 0.00 0.02 0.00 0.00 0.00 0.00 0.08 0.16 0.16 0.08
TM-FFA-CA 0.15 0.14 0.01 0.06 0.21 0.87 0.00 0.76 0.01 0.02 0.12 0.01 0.00 0.04 0.03 0.00 0.00 0.45 0.63 0.12 0.18
TM-FFA-IL 0.03 0.06 0.03 0.02 0.02 0.42 0.00 0.76 0.00 0.00 0.09 0.00 0.03 0.04 0.00 0.00 0.00 0.00 0.01 0.04 0.08
TM-FFA-IC 0.05 0.14 0.00 0.04 0.15 0.09 0.01 0.66 0.01 0.00 0.05 0.01 0.00 0.01 0.00 0.00 0.00 0.29 0.44 0.01 0.10
All 0.14 0.10 0.07 0.06 0.10 0.35 0.03 0.76 0.01 0.01 0.07 0.01 0.02 0.01 0.14 0.05 0.04 0.15 0.19 0.06 0.12
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 14/27
![Page 119: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/119.jpg)
Setup po2 su2 po3 su3 fac gcd exp ℓ20 prm ssq sra srb ild lsb mul qad mod mi5 ma5 sm5 All
DIR 0.10 0.10 0.05 0.05 0.07 0.29 0.03 0.66 0.00 0.02 0.06 0.00 0.03 0.01 0.12 0.01 0.03 0.10 0.13 0.06 0.10
FFA 0.18 0.11 0.09 0.07 0.13 0.41 0.03 0.86 0.01 0.01 0.08 0.02 0.02 0.01 0.17 0.09 0.05 0.20 0.24 0.07 0.14
SM 0.16 0.12 0.09 0.07 0.07 0.48 0.05 0.87 0.01 0.02 0.08 0.02 0.03 0.01 0.23 0.07 0.07 0.20 0.23 0.04 0.15
TM 0.12 0.09 0.05 0.05 0.13 0.22 0.00 0.65 0.00 0.01 0.06 0.00 0.02 0.01 0.06 0.04 0.01 0.09 0.14 0.09 0.09
SM-DIR-CL 0.14 0.03 0.04 0.01 0.01 0.37 0.20 0.93 0.00 0.00 0.01 0.00 0.00 0.00 0.44 0.07 0.22 0.04 0.07 0.11 0.13
SM-DIR-ML 0.18 0.66 0.08 0.32 0.27 0.36 0.11 0.70 0.00 0.21 0.03 0.00 0.01 0.00 0.63 0.01 0.13 0.00 0.00 0.04 0.19
SM-DIR-WL 0.11 0.00 0.08 0.00 0.00 0.42 0.00 0.51 0.00 0.00 0.01 0.00 0.00 0.01 0.00 0.00 0.00 0.00 0.00 0.07 0.06
SM-DIR-CA 0.11 0.00 0.08 0.00 0.00 0.55 0.00 0.93 0.01 0.00 0.14 0.01 0.13 0.03 0.00 0.00 0.00 0.40 0.50 0.00 0.14
SM-DIR-IL 0.09 0.01 0.11 0.01 0.00 0.10 0.00 0.81 0.00 0.00 0.04 0.00 0.01 0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.06
SM-DIR-IC 0.04 0.03 0.00 0.00 0.00 0.68 0.00 0.88 0.01 0.00 0.13 0.02 0.05 0.02 0.01 0.00 0.00 0.43 0.55 0.02 0.14
SM-FFA-CL 0.32 0.03 0.16 0.00 0.03 0.41 0.19 1.00 0.00 0.00 0.01 0.07 0.00 0.00 0.72 0.41 0.31 0.19 0.17 0.05 0.20
SM-FFA-ML 0.14 0.57 0.15 0.46 0.45 0.47 0.12 0.92 0.02 0.05 0.05 0.00 0.00 0.00 0.89 0.30 0.15 0.00 0.00 0.06 0.24
SM-FFA-WL 0.32 0.00 0.17 0.02 0.04 0.46 0.00 0.90 0.01 0.00 0.01 0.01 0.05 0.01 0.00 0.00 0.00 0.00 0.02 0.06 0.10
SM-FFA-CA 0.17 0.00 0.05 0.00 0.02 0.80 0.00 0.97 0.02 0.00 0.22 0.06 0.02 0.01 0.00 0.00 0.00 0.64 0.74 0.00 0.19
SM-FFA-IL 0.21 0.03 0.11 0.00 0.00 0.42 0.01 0.93 0.01 0.00 0.15 0.01 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.05 0.10
SM-FFA-IC 0.12 0.02 0.01 0.03 0.00 0.76 0.00 0.90 0.05 0.00 0.16 0.03 0.01 0.01 0.01 0.00 0.00 0.72 0.74 0.04 0.18
TM-DIR-CL 0.12 0.07 0.07 0.03 0.01 0.00 0.00 0.81 0.00 0.00 0.00 0.00 0.00 0.00 0.26 0.05 0.04 0.00 0.00 0.13 0.08
TM-DIR-ML 0.17 0.27 0.04 0.14 0.43 0.01 0.00 0.30 0.00 0.01 0.18 0.00 0.03 0.00 0.03 0.00 0.01 0.00 0.00 0.14 0.09
TM-DIR-WL 0.15 0.00 0.07 0.00 0.00 0.11 0.00 0.24 0.00 0.00 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.01 0.03 0.08 0.04
TM-DIR-CA 0.04 0.05 0.04 0.08 0.05 0.83 0.00 0.66 0.00 0.00 0.11 0.01 0.06 0.01 0.04 0.01 0.00 0.24 0.30 0.10 0.13
TM-DIR-IL 0.03 0.01 0.02 0.00 0.00 0.04 0.00 0.75 0.00 0.00 0.02 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.01 0.00 0.05
TM-DIR-IC 0.04 0.03 0.00 0.01 0.09 0.05 0.00 0.45 0.00 0.00 0.04 0.00 0.03 0.02 0.00 0.00 0.00 0.05 0.15 0.06 0.05
TM-FFA-CL 0.20 0.04 0.12 0.06 0.01 0.00 0.02 1.00 0.00 0.00 0.00 0.00 0.01 0.00 0.34 0.32 0.07 0.00 0.00 0.09 0.11
TM-FFA-ML 0.24 0.28 0.09 0.13 0.58 0.00 0.00 0.72 0.00 0.03 0.05 0.00 0.01 0.00 0.07 0.09 0.04 0.00 0.00 0.12 0.12
TM-FFA-WL 0.17 0.00 0.15 0.00 0.00 0.17 0.00 0.74 0.01 0.00 0.01 0.00 0.02 0.00 0.00 0.00 0.00 0.08 0.16 0.16 0.08
TM-FFA-CA 0.15 0.14 0.01 0.06 0.21 0.87 0.00 0.76 0.01 0.02 0.12 0.01 0.00 0.04 0.03 0.00 0.00 0.45 0.63 0.12 0.18
TM-FFA-IL 0.03 0.06 0.03 0.02 0.02 0.42 0.00 0.76 0.00 0.00 0.09 0.00 0.03 0.04 0.00 0.00 0.00 0.00 0.01 0.04 0.08
TM-FFA-IC 0.05 0.14 0.00 0.04 0.15 0.09 0.01 0.66 0.01 0.00 0.05 0.01 0.00 0.01 0.00 0.00 0.00 0.29 0.44 0.01 0.10
All 0.14 0.10 0.07 0.06 0.10 0.35 0.03 0.76 0.01 0.01 0.07 0.01 0.02 0.01 0.14 0.05 0.04 0.15 0.19 0.06 0.12
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 14/27
benchmarks
![Page 120: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/120.jpg)
Setup po2 su2 po3 su3 fac gcd exp ℓ20 prm ssq sra srb ild lsb mul qad mod mi5 ma5 sm5 All
DIR 0.10 0.10 0.05 0.05 0.07 0.29 0.03 0.66 0.00 0.02 0.06 0.00 0.03 0.01 0.12 0.01 0.03 0.10 0.13 0.06 0.10
FFA 0.18 0.11 0.09 0.07 0.13 0.41 0.03 0.86 0.01 0.01 0.08 0.02 0.02 0.01 0.17 0.09 0.05 0.20 0.24 0.07 0.14
SM 0.16 0.12 0.09 0.07 0.07 0.48 0.05 0.87 0.01 0.02 0.08 0.02 0.03 0.01 0.23 0.07 0.07 0.20 0.23 0.04 0.15
TM 0.12 0.09 0.05 0.05 0.13 0.22 0.00 0.65 0.00 0.01 0.06 0.00 0.02 0.01 0.06 0.04 0.01 0.09 0.14 0.09 0.09
SM-DIR-CL 0.14 0.03 0.04 0.01 0.01 0.37 0.20 0.93 0.00 0.00 0.01 0.00 0.00 0.00 0.44 0.07 0.22 0.04 0.07 0.11 0.13
SM-DIR-ML 0.18 0.66 0.08 0.32 0.27 0.36 0.11 0.70 0.00 0.21 0.03 0.00 0.01 0.00 0.63 0.01 0.13 0.00 0.00 0.04 0.19
SM-DIR-WL 0.11 0.00 0.08 0.00 0.00 0.42 0.00 0.51 0.00 0.00 0.01 0.00 0.00 0.01 0.00 0.00 0.00 0.00 0.00 0.07 0.06
SM-DIR-CA 0.11 0.00 0.08 0.00 0.00 0.55 0.00 0.93 0.01 0.00 0.14 0.01 0.13 0.03 0.00 0.00 0.00 0.40 0.50 0.00 0.14
SM-DIR-IL 0.09 0.01 0.11 0.01 0.00 0.10 0.00 0.81 0.00 0.00 0.04 0.00 0.01 0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.06
SM-DIR-IC 0.04 0.03 0.00 0.00 0.00 0.68 0.00 0.88 0.01 0.00 0.13 0.02 0.05 0.02 0.01 0.00 0.00 0.43 0.55 0.02 0.14
SM-FFA-CL 0.32 0.03 0.16 0.00 0.03 0.41 0.19 1.00 0.00 0.00 0.01 0.07 0.00 0.00 0.72 0.41 0.31 0.19 0.17 0.05 0.20
SM-FFA-ML 0.14 0.57 0.15 0.46 0.45 0.47 0.12 0.92 0.02 0.05 0.05 0.00 0.00 0.00 0.89 0.30 0.15 0.00 0.00 0.06 0.24
SM-FFA-WL 0.32 0.00 0.17 0.02 0.04 0.46 0.00 0.90 0.01 0.00 0.01 0.01 0.05 0.01 0.00 0.00 0.00 0.00 0.02 0.06 0.10
SM-FFA-CA 0.17 0.00 0.05 0.00 0.02 0.80 0.00 0.97 0.02 0.00 0.22 0.06 0.02 0.01 0.00 0.00 0.00 0.64 0.74 0.00 0.19
SM-FFA-IL 0.21 0.03 0.11 0.00 0.00 0.42 0.01 0.93 0.01 0.00 0.15 0.01 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.05 0.10
SM-FFA-IC 0.12 0.02 0.01 0.03 0.00 0.76 0.00 0.90 0.05 0.00 0.16 0.03 0.01 0.01 0.01 0.00 0.00 0.72 0.74 0.04 0.18
TM-DIR-CL 0.12 0.07 0.07 0.03 0.01 0.00 0.00 0.81 0.00 0.00 0.00 0.00 0.00 0.00 0.26 0.05 0.04 0.00 0.00 0.13 0.08
TM-DIR-ML 0.17 0.27 0.04 0.14 0.43 0.01 0.00 0.30 0.00 0.01 0.18 0.00 0.03 0.00 0.03 0.00 0.01 0.00 0.00 0.14 0.09
TM-DIR-WL 0.15 0.00 0.07 0.00 0.00 0.11 0.00 0.24 0.00 0.00 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.01 0.03 0.08 0.04
TM-DIR-CA 0.04 0.05 0.04 0.08 0.05 0.83 0.00 0.66 0.00 0.00 0.11 0.01 0.06 0.01 0.04 0.01 0.00 0.24 0.30 0.10 0.13
TM-DIR-IL 0.03 0.01 0.02 0.00 0.00 0.04 0.00 0.75 0.00 0.00 0.02 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.01 0.00 0.05
TM-DIR-IC 0.04 0.03 0.00 0.01 0.09 0.05 0.00 0.45 0.00 0.00 0.04 0.00 0.03 0.02 0.00 0.00 0.00 0.05 0.15 0.06 0.05
TM-FFA-CL 0.20 0.04 0.12 0.06 0.01 0.00 0.02 1.00 0.00 0.00 0.00 0.00 0.01 0.00 0.34 0.32 0.07 0.00 0.00 0.09 0.11
TM-FFA-ML 0.24 0.28 0.09 0.13 0.58 0.00 0.00 0.72 0.00 0.03 0.05 0.00 0.01 0.00 0.07 0.09 0.04 0.00 0.00 0.12 0.12
TM-FFA-WL 0.17 0.00 0.15 0.00 0.00 0.17 0.00 0.74 0.01 0.00 0.01 0.00 0.02 0.00 0.00 0.00 0.00 0.08 0.16 0.16 0.08
TM-FFA-CA 0.15 0.14 0.01 0.06 0.21 0.87 0.00 0.76 0.01 0.02 0.12 0.01 0.00 0.04 0.03 0.00 0.00 0.45 0.63 0.12 0.18
TM-FFA-IL 0.03 0.06 0.03 0.02 0.02 0.42 0.00 0.76 0.00 0.00 0.09 0.00 0.03 0.04 0.00 0.00 0.00 0.00 0.01 0.04 0.08
TM-FFA-IC 0.05 0.14 0.00 0.04 0.15 0.09 0.01 0.66 0.01 0.00 0.05 0.01 0.00 0.01 0.00 0.00 0.00 0.29 0.44 0.01 0.10
All 0.14 0.10 0.07 0.06 0.10 0.35 0.03 0.76 0.01 0.01 0.07 0.01 0.02 0.01 0.14 0.05 0.04 0.15 0.19 0.06 0.12
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 14/27
benchmarks
setups
![Page 121: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/121.jpg)
Setup po2 su2 po3 su3 fac gcd exp ℓ20 prm ssq sra srb ild lsb mul qad mod mi5 ma5 sm5 All
DIR 0.10 0.10 0.05 0.05 0.07 0.29 0.03 0.66 0.00 0.02 0.06 0.00 0.03 0.01 0.12 0.01 0.03 0.10 0.13 0.06 0.10
FFA 0.18 0.11 0.09 0.07 0.13 0.41 0.03 0.86 0.01 0.01 0.08 0.02 0.02 0.01 0.17 0.09 0.05 0.20 0.24 0.07 0.14
SM 0.16 0.12 0.09 0.07 0.07 0.48 0.05 0.87 0.01 0.02 0.08 0.02 0.03 0.01 0.23 0.07 0.07 0.20 0.23 0.04 0.15
TM 0.12 0.09 0.05 0.05 0.13 0.22 0.00 0.65 0.00 0.01 0.06 0.00 0.02 0.01 0.06 0.04 0.01 0.09 0.14 0.09 0.09
SM-DIR-CL 0.14 0.03 0.04 0.01 0.01 0.37 0.20 0.93 0.00 0.00 0.01 0.00 0.00 0.00 0.44 0.07 0.22 0.04 0.07 0.11 0.13
SM-DIR-ML 0.18 0.66 0.08 0.32 0.27 0.36 0.11 0.70 0.00 0.21 0.03 0.00 0.01 0.00 0.63 0.01 0.13 0.00 0.00 0.04 0.19
SM-DIR-WL 0.11 0.00 0.08 0.00 0.00 0.42 0.00 0.51 0.00 0.00 0.01 0.00 0.00 0.01 0.00 0.00 0.00 0.00 0.00 0.07 0.06
SM-DIR-CA 0.11 0.00 0.08 0.00 0.00 0.55 0.00 0.93 0.01 0.00 0.14 0.01 0.13 0.03 0.00 0.00 0.00 0.40 0.50 0.00 0.14
SM-DIR-IL 0.09 0.01 0.11 0.01 0.00 0.10 0.00 0.81 0.00 0.00 0.04 0.00 0.01 0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.06
SM-DIR-IC 0.04 0.03 0.00 0.00 0.00 0.68 0.00 0.88 0.01 0.00 0.13 0.02 0.05 0.02 0.01 0.00 0.00 0.43 0.55 0.02 0.14
SM-FFA-CL 0.32 0.03 0.16 0.00 0.03 0.41 0.19 1.00 0.00 0.00 0.01 0.07 0.00 0.00 0.72 0.41 0.31 0.19 0.17 0.05 0.20
SM-FFA-ML 0.14 0.57 0.15 0.46 0.45 0.47 0.12 0.92 0.02 0.05 0.05 0.00 0.00 0.00 0.89 0.30 0.15 0.00 0.00 0.06 0.24
SM-FFA-WL 0.32 0.00 0.17 0.02 0.04 0.46 0.00 0.90 0.01 0.00 0.01 0.01 0.05 0.01 0.00 0.00 0.00 0.00 0.02 0.06 0.10
SM-FFA-CA 0.17 0.00 0.05 0.00 0.02 0.80 0.00 0.97 0.02 0.00 0.22 0.06 0.02 0.01 0.00 0.00 0.00 0.64 0.74 0.00 0.19
SM-FFA-IL 0.21 0.03 0.11 0.00 0.00 0.42 0.01 0.93 0.01 0.00 0.15 0.01 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.05 0.10
SM-FFA-IC 0.12 0.02 0.01 0.03 0.00 0.76 0.00 0.90 0.05 0.00 0.16 0.03 0.01 0.01 0.01 0.00 0.00 0.72 0.74 0.04 0.18
TM-DIR-CL 0.12 0.07 0.07 0.03 0.01 0.00 0.00 0.81 0.00 0.00 0.00 0.00 0.00 0.00 0.26 0.05 0.04 0.00 0.00 0.13 0.08
TM-DIR-ML 0.17 0.27 0.04 0.14 0.43 0.01 0.00 0.30 0.00 0.01 0.18 0.00 0.03 0.00 0.03 0.00 0.01 0.00 0.00 0.14 0.09
TM-DIR-WL 0.15 0.00 0.07 0.00 0.00 0.11 0.00 0.24 0.00 0.00 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.01 0.03 0.08 0.04
TM-DIR-CA 0.04 0.05 0.04 0.08 0.05 0.83 0.00 0.66 0.00 0.00 0.11 0.01 0.06 0.01 0.04 0.01 0.00 0.24 0.30 0.10 0.13
TM-DIR-IL 0.03 0.01 0.02 0.00 0.00 0.04 0.00 0.75 0.00 0.00 0.02 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.01 0.00 0.05
TM-DIR-IC 0.04 0.03 0.00 0.01 0.09 0.05 0.00 0.45 0.00 0.00 0.04 0.00 0.03 0.02 0.00 0.00 0.00 0.05 0.15 0.06 0.05
TM-FFA-CL 0.20 0.04 0.12 0.06 0.01 0.00 0.02 1.00 0.00 0.00 0.00 0.00 0.01 0.00 0.34 0.32 0.07 0.00 0.00 0.09 0.11
TM-FFA-ML 0.24 0.28 0.09 0.13 0.58 0.00 0.00 0.72 0.00 0.03 0.05 0.00 0.01 0.00 0.07 0.09 0.04 0.00 0.00 0.12 0.12
TM-FFA-WL 0.17 0.00 0.15 0.00 0.00 0.17 0.00 0.74 0.01 0.00 0.01 0.00 0.02 0.00 0.00 0.00 0.00 0.08 0.16 0.16 0.08
TM-FFA-CA 0.15 0.14 0.01 0.06 0.21 0.87 0.00 0.76 0.01 0.02 0.12 0.01 0.00 0.04 0.03 0.00 0.00 0.45 0.63 0.12 0.18
TM-FFA-IL 0.03 0.06 0.03 0.02 0.02 0.42 0.00 0.76 0.00 0.00 0.09 0.00 0.03 0.04 0.00 0.00 0.00 0.00 0.01 0.04 0.08
TM-FFA-IC 0.05 0.14 0.00 0.04 0.15 0.09 0.01 0.66 0.01 0.00 0.05 0.01 0.00 0.01 0.00 0.00 0.00 0.29 0.44 0.01 0.10
All 0.14 0.10 0.07 0.06 0.10 0.35 0.03 0.76 0.01 0.01 0.07 0.01 0.02 0.01 0.14 0.05 0.04 0.15 0.19 0.06 0.12
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 14/27
benchmarks
setups
aggregated results:
with FFA (and f directly DIR)
with TM (and simple memory SM)
![Page 122: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/122.jpg)
Setup po2 su2 po3 su3 fac gcd exp ℓ20 prm ssq sra srb ild lsb mul qad mod mi5 ma5 sm5 All
DIR 0.10 0.10 0.05 0.05 0.07 0.29 0.03 0.66 0.00 0.02 0.06 0.00 0.03 0.01 0.12 0.01 0.03 0.10 0.13 0.06 0.10
FFA 0.18 0.11 0.09 0.07 0.13 0.41 0.03 0.86 0.01 0.01 0.08 0.02 0.02 0.01 0.17 0.09 0.05 0.20 0.24 0.07 0.14
SM 0.16 0.12 0.09 0.07 0.07 0.48 0.05 0.87 0.01 0.02 0.08 0.02 0.03 0.01 0.23 0.07 0.07 0.20 0.23 0.04 0.15
TM 0.12 0.09 0.05 0.05 0.13 0.22 0.00 0.65 0.00 0.01 0.06 0.00 0.02 0.01 0.06 0.04 0.01 0.09 0.14 0.09 0.09
SM-DIR-CL 0.14 0.03 0.04 0.01 0.01 0.37 0.20 0.93 0.00 0.00 0.01 0.00 0.00 0.00 0.44 0.07 0.22 0.04 0.07 0.11 0.13
SM-DIR-ML 0.18 0.66 0.08 0.32 0.27 0.36 0.11 0.70 0.00 0.21 0.03 0.00 0.01 0.00 0.63 0.01 0.13 0.00 0.00 0.04 0.19
SM-DIR-WL 0.11 0.00 0.08 0.00 0.00 0.42 0.00 0.51 0.00 0.00 0.01 0.00 0.00 0.01 0.00 0.00 0.00 0.00 0.00 0.07 0.06
SM-DIR-CA 0.11 0.00 0.08 0.00 0.00 0.55 0.00 0.93 0.01 0.00 0.14 0.01 0.13 0.03 0.00 0.00 0.00 0.40 0.50 0.00 0.14
SM-DIR-IL 0.09 0.01 0.11 0.01 0.00 0.10 0.00 0.81 0.00 0.00 0.04 0.00 0.01 0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.06
SM-DIR-IC 0.04 0.03 0.00 0.00 0.00 0.68 0.00 0.88 0.01 0.00 0.13 0.02 0.05 0.02 0.01 0.00 0.00 0.43 0.55 0.02 0.14
SM-FFA-CL 0.32 0.03 0.16 0.00 0.03 0.41 0.19 1.00 0.00 0.00 0.01 0.07 0.00 0.00 0.72 0.41 0.31 0.19 0.17 0.05 0.20
SM-FFA-ML 0.14 0.57 0.15 0.46 0.45 0.47 0.12 0.92 0.02 0.05 0.05 0.00 0.00 0.00 0.89 0.30 0.15 0.00 0.00 0.06 0.24
SM-FFA-WL 0.32 0.00 0.17 0.02 0.04 0.46 0.00 0.90 0.01 0.00 0.01 0.01 0.05 0.01 0.00 0.00 0.00 0.00 0.02 0.06 0.10
SM-FFA-CA 0.17 0.00 0.05 0.00 0.02 0.80 0.00 0.97 0.02 0.00 0.22 0.06 0.02 0.01 0.00 0.00 0.00 0.64 0.74 0.00 0.19
SM-FFA-IL 0.21 0.03 0.11 0.00 0.00 0.42 0.01 0.93 0.01 0.00 0.15 0.01 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.05 0.10
SM-FFA-IC 0.12 0.02 0.01 0.03 0.00 0.76 0.00 0.90 0.05 0.00 0.16 0.03 0.01 0.01 0.01 0.00 0.00 0.72 0.74 0.04 0.18
TM-DIR-CL 0.12 0.07 0.07 0.03 0.01 0.00 0.00 0.81 0.00 0.00 0.00 0.00 0.00 0.00 0.26 0.05 0.04 0.00 0.00 0.13 0.08
TM-DIR-ML 0.17 0.27 0.04 0.14 0.43 0.01 0.00 0.30 0.00 0.01 0.18 0.00 0.03 0.00 0.03 0.00 0.01 0.00 0.00 0.14 0.09
TM-DIR-WL 0.15 0.00 0.07 0.00 0.00 0.11 0.00 0.24 0.00 0.00 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.01 0.03 0.08 0.04
TM-DIR-CA 0.04 0.05 0.04 0.08 0.05 0.83 0.00 0.66 0.00 0.00 0.11 0.01 0.06 0.01 0.04 0.01 0.00 0.24 0.30 0.10 0.13
TM-DIR-IL 0.03 0.01 0.02 0.00 0.00 0.04 0.00 0.75 0.00 0.00 0.02 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.01 0.00 0.05
TM-DIR-IC 0.04 0.03 0.00 0.01 0.09 0.05 0.00 0.45 0.00 0.00 0.04 0.00 0.03 0.02 0.00 0.00 0.00 0.05 0.15 0.06 0.05
TM-FFA-CL 0.20 0.04 0.12 0.06 0.01 0.00 0.02 1.00 0.00 0.00 0.00 0.00 0.01 0.00 0.34 0.32 0.07 0.00 0.00 0.09 0.11
TM-FFA-ML 0.24 0.28 0.09 0.13 0.58 0.00 0.00 0.72 0.00 0.03 0.05 0.00 0.01 0.00 0.07 0.09 0.04 0.00 0.00 0.12 0.12
TM-FFA-WL 0.17 0.00 0.15 0.00 0.00 0.17 0.00 0.74 0.01 0.00 0.01 0.00 0.02 0.00 0.00 0.00 0.00 0.08 0.16 0.16 0.08
TM-FFA-CA 0.15 0.14 0.01 0.06 0.21 0.87 0.00 0.76 0.01 0.02 0.12 0.01 0.00 0.04 0.03 0.00 0.00 0.45 0.63 0.12 0.18
TM-FFA-IL 0.03 0.06 0.03 0.02 0.02 0.42 0.00 0.76 0.00 0.00 0.09 0.00 0.03 0.04 0.00 0.00 0.00 0.00 0.01 0.04 0.08
TM-FFA-IC 0.05 0.14 0.00 0.04 0.15 0.09 0.01 0.66 0.01 0.00 0.05 0.01 0.00 0.01 0.00 0.00 0.00 0.29 0.44 0.01 0.10
All 0.14 0.10 0.07 0.06 0.10 0.35 0.03 0.76 0.01 0.01 0.07 0.01 0.02 0.01 0.14 0.05 0.04 0.15 0.19 0.06 0.12
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 14/27
benchmarks
setups
aggregated results:
with FFA (and f directly DIR)
with TM (and simple memory SM)
aggregated results:
over benchmark instances
over setups
![Page 123: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/123.jpg)
Setup po2 su2 po3 su3 fac gcd exp ℓ20 prm ssq sra srb ild lsb mul qad mod mi5 ma5 sm5 All
DIR 0.10 0.10 0.05 0.05 0.07 0.29 0.03 0.66 0.00 0.02 0.06 0.00 0.03 0.01 0.12 0.01 0.03 0.10 0.13 0.06 0.10
FFA 0.18 0.11 0.09 0.07 0.13 0.41 0.03 0.86 0.01 0.01 0.08 0.02 0.02 0.01 0.17 0.09 0.05 0.20 0.24 0.07 0.14
SM 0.16 0.12 0.09 0.07 0.07 0.48 0.05 0.87 0.01 0.02 0.08 0.02 0.03 0.01 0.23 0.07 0.07 0.20 0.23 0.04 0.15
TM 0.12 0.09 0.05 0.05 0.13 0.22 0.00 0.65 0.00 0.01 0.06 0.00 0.02 0.01 0.06 0.04 0.01 0.09 0.14 0.09 0.09
SM-DIR-CL 0.14 0.03 0.04 0.01 0.01 0.37 0.20 0.93 0.00 0.00 0.01 0.00 0.00 0.00 0.44 0.07 0.22 0.04 0.07 0.11 0.13
SM-DIR-ML 0.18 0.66 0.08 0.32 0.27 0.36 0.11 0.70 0.00 0.21 0.03 0.00 0.01 0.00 0.63 0.01 0.13 0.00 0.00 0.04 0.19
SM-DIR-WL 0.11 0.00 0.08 0.00 0.00 0.42 0.00 0.51 0.00 0.00 0.01 0.00 0.00 0.01 0.00 0.00 0.00 0.00 0.00 0.07 0.06
SM-DIR-CA 0.11 0.00 0.08 0.00 0.00 0.55 0.00 0.93 0.01 0.00 0.14 0.01 0.13 0.03 0.00 0.00 0.00 0.40 0.50 0.00 0.14
SM-DIR-IL 0.09 0.01 0.11 0.01 0.00 0.10 0.00 0.81 0.00 0.00 0.04 0.00 0.01 0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.06
SM-DIR-IC 0.04 0.03 0.00 0.00 0.00 0.68 0.00 0.88 0.01 0.00 0.13 0.02 0.05 0.02 0.01 0.00 0.00 0.43 0.55 0.02 0.14
SM-FFA-CL 0.32 0.03 0.16 0.00 0.03 0.41 0.19 1.00 0.00 0.00 0.01 0.07 0.00 0.00 0.72 0.41 0.31 0.19 0.17 0.05 0.20
SM-FFA-ML 0.14 0.57 0.15 0.46 0.45 0.47 0.12 0.92 0.02 0.05 0.05 0.00 0.00 0.00 0.89 0.30 0.15 0.00 0.00 0.06 0.24
SM-FFA-WL 0.32 0.00 0.17 0.02 0.04 0.46 0.00 0.90 0.01 0.00 0.01 0.01 0.05 0.01 0.00 0.00 0.00 0.00 0.02 0.06 0.10
SM-FFA-CA 0.17 0.00 0.05 0.00 0.02 0.80 0.00 0.97 0.02 0.00 0.22 0.06 0.02 0.01 0.00 0.00 0.00 0.64 0.74 0.00 0.19
SM-FFA-IL 0.21 0.03 0.11 0.00 0.00 0.42 0.01 0.93 0.01 0.00 0.15 0.01 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.05 0.10
SM-FFA-IC 0.12 0.02 0.01 0.03 0.00 0.76 0.00 0.90 0.05 0.00 0.16 0.03 0.01 0.01 0.01 0.00 0.00 0.72 0.74 0.04 0.18
TM-DIR-CL 0.12 0.07 0.07 0.03 0.01 0.00 0.00 0.81 0.00 0.00 0.00 0.00 0.00 0.00 0.26 0.05 0.04 0.00 0.00 0.13 0.08
TM-DIR-ML 0.17 0.27 0.04 0.14 0.43 0.01 0.00 0.30 0.00 0.01 0.18 0.00 0.03 0.00 0.03 0.00 0.01 0.00 0.00 0.14 0.09
TM-DIR-WL 0.15 0.00 0.07 0.00 0.00 0.11 0.00 0.24 0.00 0.00 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.01 0.03 0.08 0.04
TM-DIR-CA 0.04 0.05 0.04 0.08 0.05 0.83 0.00 0.66 0.00 0.00 0.11 0.01 0.06 0.01 0.04 0.01 0.00 0.24 0.30 0.10 0.13
TM-DIR-IL 0.03 0.01 0.02 0.00 0.00 0.04 0.00 0.75 0.00 0.00 0.02 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.01 0.00 0.05
TM-DIR-IC 0.04 0.03 0.00 0.01 0.09 0.05 0.00 0.45 0.00 0.00 0.04 0.00 0.03 0.02 0.00 0.00 0.00 0.05 0.15 0.06 0.05
TM-FFA-CL 0.20 0.04 0.12 0.06 0.01 0.00 0.02 1.00 0.00 0.00 0.00 0.00 0.01 0.00 0.34 0.32 0.07 0.00 0.00 0.09 0.11
TM-FFA-ML 0.24 0.28 0.09 0.13 0.58 0.00 0.00 0.72 0.00 0.03 0.05 0.00 0.01 0.00 0.07 0.09 0.04 0.00 0.00 0.12 0.12
TM-FFA-WL 0.17 0.00 0.15 0.00 0.00 0.17 0.00 0.74 0.01 0.00 0.01 0.00 0.02 0.00 0.00 0.00 0.00 0.08 0.16 0.16 0.08
TM-FFA-CA 0.15 0.14 0.01 0.06 0.21 0.87 0.00 0.76 0.01 0.02 0.12 0.01 0.00 0.04 0.03 0.00 0.00 0.45 0.63 0.12 0.18
TM-FFA-IL 0.03 0.06 0.03 0.02 0.02 0.42 0.00 0.76 0.00 0.00 0.09 0.00 0.03 0.04 0.00 0.00 0.00 0.00 0.01 0.04 0.08
TM-FFA-IC 0.05 0.14 0.00 0.04 0.15 0.09 0.01 0.66 0.01 0.00 0.05 0.01 0.00 0.01 0.00 0.00 0.00 0.29 0.44 0.01 0.10
All 0.14 0.10 0.07 0.06 0.10 0.35 0.03 0.76 0.01 0.01 0.07 0.01 0.02 0.01 0.14 0.05 0.04 0.15 0.19 0.06 0.12
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 14/27
FFA increases success rate
by 40%
![Page 124: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/124.jpg)
Setup po2 su2 po3 su3 fac gcd exp ℓ20 prm ssq sra srb ild lsb mul qad mod mi5 ma5 sm5 All
DIR 0.10 0.10 0.05 0.05 0.07 0.29 0.03 0.66 0.00 0.02 0.06 0.00 0.03 0.01 0.12 0.01 0.03 0.10 0.13 0.06 0.10
FFA 0.18 0.11 0.09 0.07 0.13 0.41 0.03 0.86 0.01 0.01 0.08 0.02 0.02 0.01 0.17 0.09 0.05 0.20 0.24 0.07 0.14
SM 0.16 0.12 0.09 0.07 0.07 0.48 0.05 0.87 0.01 0.02 0.08 0.02 0.03 0.01 0.23 0.07 0.07 0.20 0.23 0.04 0.15
TM 0.12 0.09 0.05 0.05 0.13 0.22 0.00 0.65 0.00 0.01 0.06 0.00 0.02 0.01 0.06 0.04 0.01 0.09 0.14 0.09 0.09
SM-DIR-CL 0.14 0.03 0.04 0.01 0.01 0.37 0.20 0.93 0.00 0.00 0.01 0.00 0.00 0.00 0.44 0.07 0.22 0.04 0.07 0.11 0.13
SM-DIR-ML 0.18 0.66 0.08 0.32 0.27 0.36 0.11 0.70 0.00 0.21 0.03 0.00 0.01 0.00 0.63 0.01 0.13 0.00 0.00 0.04 0.19
SM-DIR-WL 0.11 0.00 0.08 0.00 0.00 0.42 0.00 0.51 0.00 0.00 0.01 0.00 0.00 0.01 0.00 0.00 0.00 0.00 0.00 0.07 0.06
SM-DIR-CA 0.11 0.00 0.08 0.00 0.00 0.55 0.00 0.93 0.01 0.00 0.14 0.01 0.13 0.03 0.00 0.00 0.00 0.40 0.50 0.00 0.14
SM-DIR-IL 0.09 0.01 0.11 0.01 0.00 0.10 0.00 0.81 0.00 0.00 0.04 0.00 0.01 0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.06
SM-DIR-IC 0.04 0.03 0.00 0.00 0.00 0.68 0.00 0.88 0.01 0.00 0.13 0.02 0.05 0.02 0.01 0.00 0.00 0.43 0.55 0.02 0.14
SM-FFA-CL 0.32 0.03 0.16 0.00 0.03 0.41 0.19 1.00 0.00 0.00 0.01 0.07 0.00 0.00 0.72 0.41 0.31 0.19 0.17 0.05 0.20
SM-FFA-ML 0.14 0.57 0.15 0.46 0.45 0.47 0.12 0.92 0.02 0.05 0.05 0.00 0.00 0.00 0.89 0.30 0.15 0.00 0.00 0.06 0.24
SM-FFA-WL 0.32 0.00 0.17 0.02 0.04 0.46 0.00 0.90 0.01 0.00 0.01 0.01 0.05 0.01 0.00 0.00 0.00 0.00 0.02 0.06 0.10
SM-FFA-CA 0.17 0.00 0.05 0.00 0.02 0.80 0.00 0.97 0.02 0.00 0.22 0.06 0.02 0.01 0.00 0.00 0.00 0.64 0.74 0.00 0.19
SM-FFA-IL 0.21 0.03 0.11 0.00 0.00 0.42 0.01 0.93 0.01 0.00 0.15 0.01 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.05 0.10
SM-FFA-IC 0.12 0.02 0.01 0.03 0.00 0.76 0.00 0.90 0.05 0.00 0.16 0.03 0.01 0.01 0.01 0.00 0.00 0.72 0.74 0.04 0.18
TM-DIR-CL 0.12 0.07 0.07 0.03 0.01 0.00 0.00 0.81 0.00 0.00 0.00 0.00 0.00 0.00 0.26 0.05 0.04 0.00 0.00 0.13 0.08
TM-DIR-ML 0.17 0.27 0.04 0.14 0.43 0.01 0.00 0.30 0.00 0.01 0.18 0.00 0.03 0.00 0.03 0.00 0.01 0.00 0.00 0.14 0.09
TM-DIR-WL 0.15 0.00 0.07 0.00 0.00 0.11 0.00 0.24 0.00 0.00 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.01 0.03 0.08 0.04
TM-DIR-CA 0.04 0.05 0.04 0.08 0.05 0.83 0.00 0.66 0.00 0.00 0.11 0.01 0.06 0.01 0.04 0.01 0.00 0.24 0.30 0.10 0.13
TM-DIR-IL 0.03 0.01 0.02 0.00 0.00 0.04 0.00 0.75 0.00 0.00 0.02 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.01 0.00 0.05
TM-DIR-IC 0.04 0.03 0.00 0.01 0.09 0.05 0.00 0.45 0.00 0.00 0.04 0.00 0.03 0.02 0.00 0.00 0.00 0.05 0.15 0.06 0.05
TM-FFA-CL 0.20 0.04 0.12 0.06 0.01 0.00 0.02 1.00 0.00 0.00 0.00 0.00 0.01 0.00 0.34 0.32 0.07 0.00 0.00 0.09 0.11
TM-FFA-ML 0.24 0.28 0.09 0.13 0.58 0.00 0.00 0.72 0.00 0.03 0.05 0.00 0.01 0.00 0.07 0.09 0.04 0.00 0.00 0.12 0.12
TM-FFA-WL 0.17 0.00 0.15 0.00 0.00 0.17 0.00 0.74 0.01 0.00 0.01 0.00 0.02 0.00 0.00 0.00 0.00 0.08 0.16 0.16 0.08
TM-FFA-CA 0.15 0.14 0.01 0.06 0.21 0.87 0.00 0.76 0.01 0.02 0.12 0.01 0.00 0.04 0.03 0.00 0.00 0.45 0.63 0.12 0.18
TM-FFA-IL 0.03 0.06 0.03 0.02 0.02 0.42 0.00 0.76 0.00 0.00 0.09 0.00 0.03 0.04 0.00 0.00 0.00 0.00 0.01 0.04 0.08
TM-FFA-IC 0.05 0.14 0.00 0.04 0.15 0.09 0.01 0.66 0.01 0.00 0.05 0.01 0.00 0.01 0.00 0.00 0.00 0.29 0.44 0.01 0.10
All 0.14 0.10 0.07 0.06 0.10 0.35 0.03 0.76 0.01 0.01 0.07 0.01 0.02 0.01 0.14 0.05 0.04 0.15 0.19 0.06 0.12
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 14/27
FFA increases success rate
by 40%
TM decreases success rate
on 17 out of 20 benchmarks
![Page 125: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/125.jpg)
Setup po2 su2 po3 su3 fac gcd exp ℓ20 prm ssq sra srb ild lsb mul qad mod mi5 ma5 sm5 All
DIR 0.10 0.10 0.05 0.05 0.07 0.29 0.03 0.66 0.00 0.02 0.06 0.00 0.03 0.01 0.12 0.01 0.03 0.10 0.13 0.06 0.10
FFA 0.18 0.11 0.09 0.07 0.13 0.41 0.03 0.86 0.01 0.01 0.08 0.02 0.02 0.01 0.17 0.09 0.05 0.20 0.24 0.07 0.14
SM 0.16 0.12 0.09 0.07 0.07 0.48 0.05 0.87 0.01 0.02 0.08 0.02 0.03 0.01 0.23 0.07 0.07 0.20 0.23 0.04 0.15
TM 0.12 0.09 0.05 0.05 0.13 0.22 0.00 0.65 0.00 0.01 0.06 0.00 0.02 0.01 0.06 0.04 0.01 0.09 0.14 0.09 0.09
SM-DIR-CL 0.14 0.03 0.04 0.01 0.01 0.37 0.20 0.93 0.00 0.00 0.01 0.00 0.00 0.00 0.44 0.07 0.22 0.04 0.07 0.11 0.13
SM-DIR-ML 0.18 0.66 0.08 0.32 0.27 0.36 0.11 0.70 0.00 0.21 0.03 0.00 0.01 0.00 0.63 0.01 0.13 0.00 0.00 0.04 0.19
SM-DIR-WL 0.11 0.00 0.08 0.00 0.00 0.42 0.00 0.51 0.00 0.00 0.01 0.00 0.00 0.01 0.00 0.00 0.00 0.00 0.00 0.07 0.06
SM-DIR-CA 0.11 0.00 0.08 0.00 0.00 0.55 0.00 0.93 0.01 0.00 0.14 0.01 0.13 0.03 0.00 0.00 0.00 0.40 0.50 0.00 0.14
SM-DIR-IL 0.09 0.01 0.11 0.01 0.00 0.10 0.00 0.81 0.00 0.00 0.04 0.00 0.01 0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.06
SM-DIR-IC 0.04 0.03 0.00 0.00 0.00 0.68 0.00 0.88 0.01 0.00 0.13 0.02 0.05 0.02 0.01 0.00 0.00 0.43 0.55 0.02 0.14
SM-FFA-CL 0.32 0.03 0.16 0.00 0.03 0.41 0.19 1.00 0.00 0.00 0.01 0.07 0.00 0.00 0.72 0.41 0.31 0.19 0.17 0.05 0.20
SM-FFA-ML 0.14 0.57 0.15 0.46 0.45 0.47 0.12 0.92 0.02 0.05 0.05 0.00 0.00 0.00 0.89 0.30 0.15 0.00 0.00 0.06 0.24
SM-FFA-WL 0.32 0.00 0.17 0.02 0.04 0.46 0.00 0.90 0.01 0.00 0.01 0.01 0.05 0.01 0.00 0.00 0.00 0.00 0.02 0.06 0.10
SM-FFA-CA 0.17 0.00 0.05 0.00 0.02 0.80 0.00 0.97 0.02 0.00 0.22 0.06 0.02 0.01 0.00 0.00 0.00 0.64 0.74 0.00 0.19
SM-FFA-IL 0.21 0.03 0.11 0.00 0.00 0.42 0.01 0.93 0.01 0.00 0.15 0.01 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.05 0.10
SM-FFA-IC 0.12 0.02 0.01 0.03 0.00 0.76 0.00 0.90 0.05 0.00 0.16 0.03 0.01 0.01 0.01 0.00 0.00 0.72 0.74 0.04 0.18
TM-DIR-CL 0.12 0.07 0.07 0.03 0.01 0.00 0.00 0.81 0.00 0.00 0.00 0.00 0.00 0.00 0.26 0.05 0.04 0.00 0.00 0.13 0.08
TM-DIR-ML 0.17 0.27 0.04 0.14 0.43 0.01 0.00 0.30 0.00 0.01 0.18 0.00 0.03 0.00 0.03 0.00 0.01 0.00 0.00 0.14 0.09
TM-DIR-WL 0.15 0.00 0.07 0.00 0.00 0.11 0.00 0.24 0.00 0.00 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.01 0.03 0.08 0.04
TM-DIR-CA 0.04 0.05 0.04 0.08 0.05 0.83 0.00 0.66 0.00 0.00 0.11 0.01 0.06 0.01 0.04 0.01 0.00 0.24 0.30 0.10 0.13
TM-DIR-IL 0.03 0.01 0.02 0.00 0.00 0.04 0.00 0.75 0.00 0.00 0.02 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.01 0.00 0.05
TM-DIR-IC 0.04 0.03 0.00 0.01 0.09 0.05 0.00 0.45 0.00 0.00 0.04 0.00 0.03 0.02 0.00 0.00 0.00 0.05 0.15 0.06 0.05
TM-FFA-CL 0.20 0.04 0.12 0.06 0.01 0.00 0.02 1.00 0.00 0.00 0.00 0.00 0.01 0.00 0.34 0.32 0.07 0.00 0.00 0.09 0.11
TM-FFA-ML 0.24 0.28 0.09 0.13 0.58 0.00 0.00 0.72 0.00 0.03 0.05 0.00 0.01 0.00 0.07 0.09 0.04 0.00 0.00 0.12 0.12
TM-FFA-WL 0.17 0.00 0.15 0.00 0.00 0.17 0.00 0.74 0.01 0.00 0.01 0.00 0.02 0.00 0.00 0.00 0.00 0.08 0.16 0.16 0.08
TM-FFA-CA 0.15 0.14 0.01 0.06 0.21 0.87 0.00 0.76 0.01 0.02 0.12 0.01 0.00 0.04 0.03 0.00 0.00 0.45 0.63 0.12 0.18
TM-FFA-IL 0.03 0.06 0.03 0.02 0.02 0.42 0.00 0.76 0.00 0.00 0.09 0.00 0.03 0.04 0.00 0.00 0.00 0.00 0.01 0.04 0.08
TM-FFA-IC 0.05 0.14 0.00 0.04 0.15 0.09 0.01 0.66 0.01 0.00 0.05 0.01 0.00 0.01 0.00 0.00 0.00 0.29 0.44 0.01 0.10
All 0.14 0.10 0.07 0.06 0.10 0.35 0.03 0.76 0.01 0.01 0.07 0.01 0.02 0.01 0.14 0.05 0.04 0.15 0.19 0.06 0.12
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 14/27
Best Setup: SM-FFA-ML
![Page 126: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/126.jpg)
Setup po2 su2 po3 su3 fac gcd exp ℓ20 prm ssq sra srb ild lsb mul qad mod mi5 ma5 sm5 All
DIR 0.10 0.10 0.05 0.05 0.07 0.29 0.03 0.66 0.00 0.02 0.06 0.00 0.03 0.01 0.12 0.01 0.03 0.10 0.13 0.06 0.10
FFA 0.18 0.11 0.09 0.07 0.13 0.41 0.03 0.86 0.01 0.01 0.08 0.02 0.02 0.01 0.17 0.09 0.05 0.20 0.24 0.07 0.14
SM 0.16 0.12 0.09 0.07 0.07 0.48 0.05 0.87 0.01 0.02 0.08 0.02 0.03 0.01 0.23 0.07 0.07 0.20 0.23 0.04 0.15
TM 0.12 0.09 0.05 0.05 0.13 0.22 0.00 0.65 0.00 0.01 0.06 0.00 0.02 0.01 0.06 0.04 0.01 0.09 0.14 0.09 0.09
SM-DIR-CL 0.14 0.03 0.04 0.01 0.01 0.37 0.20 0.93 0.00 0.00 0.01 0.00 0.00 0.00 0.44 0.07 0.22 0.04 0.07 0.11 0.13
SM-DIR-ML 0.18 0.66 0.08 0.32 0.27 0.36 0.11 0.70 0.00 0.21 0.03 0.00 0.01 0.00 0.63 0.01 0.13 0.00 0.00 0.04 0.19
SM-DIR-WL 0.11 0.00 0.08 0.00 0.00 0.42 0.00 0.51 0.00 0.00 0.01 0.00 0.00 0.01 0.00 0.00 0.00 0.00 0.00 0.07 0.06
SM-DIR-CA 0.11 0.00 0.08 0.00 0.00 0.55 0.00 0.93 0.01 0.00 0.14 0.01 0.13 0.03 0.00 0.00 0.00 0.40 0.50 0.00 0.14
SM-DIR-IL 0.09 0.01 0.11 0.01 0.00 0.10 0.00 0.81 0.00 0.00 0.04 0.00 0.01 0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.06
SM-DIR-IC 0.04 0.03 0.00 0.00 0.00 0.68 0.00 0.88 0.01 0.00 0.13 0.02 0.05 0.02 0.01 0.00 0.00 0.43 0.55 0.02 0.14
SM-FFA-CL 0.32 0.03 0.16 0.00 0.03 0.41 0.19 1.00 0.00 0.00 0.01 0.07 0.00 0.00 0.72 0.41 0.31 0.19 0.17 0.05 0.20
SM-FFA-ML 0.14 0.57 0.15 0.46 0.45 0.47 0.12 0.92 0.02 0.05 0.05 0.00 0.00 0.00 0.89 0.30 0.15 0.00 0.00 0.06 0.24
SM-FFA-WL 0.32 0.00 0.17 0.02 0.04 0.46 0.00 0.90 0.01 0.00 0.01 0.01 0.05 0.01 0.00 0.00 0.00 0.00 0.02 0.06 0.10
SM-FFA-CA 0.17 0.00 0.05 0.00 0.02 0.80 0.00 0.97 0.02 0.00 0.22 0.06 0.02 0.01 0.00 0.00 0.00 0.64 0.74 0.00 0.19
SM-FFA-IL 0.21 0.03 0.11 0.00 0.00 0.42 0.01 0.93 0.01 0.00 0.15 0.01 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.05 0.10
SM-FFA-IC 0.12 0.02 0.01 0.03 0.00 0.76 0.00 0.90 0.05 0.00 0.16 0.03 0.01 0.01 0.01 0.00 0.00 0.72 0.74 0.04 0.18
TM-DIR-CL 0.12 0.07 0.07 0.03 0.01 0.00 0.00 0.81 0.00 0.00 0.00 0.00 0.00 0.00 0.26 0.05 0.04 0.00 0.00 0.13 0.08
TM-DIR-ML 0.17 0.27 0.04 0.14 0.43 0.01 0.00 0.30 0.00 0.01 0.18 0.00 0.03 0.00 0.03 0.00 0.01 0.00 0.00 0.14 0.09
TM-DIR-WL 0.15 0.00 0.07 0.00 0.00 0.11 0.00 0.24 0.00 0.00 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.01 0.03 0.08 0.04
TM-DIR-CA 0.04 0.05 0.04 0.08 0.05 0.83 0.00 0.66 0.00 0.00 0.11 0.01 0.06 0.01 0.04 0.01 0.00 0.24 0.30 0.10 0.13
TM-DIR-IL 0.03 0.01 0.02 0.00 0.00 0.04 0.00 0.75 0.00 0.00 0.02 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.01 0.00 0.05
TM-DIR-IC 0.04 0.03 0.00 0.01 0.09 0.05 0.00 0.45 0.00 0.00 0.04 0.00 0.03 0.02 0.00 0.00 0.00 0.05 0.15 0.06 0.05
TM-FFA-CL 0.20 0.04 0.12 0.06 0.01 0.00 0.02 1.00 0.00 0.00 0.00 0.00 0.01 0.00 0.34 0.32 0.07 0.00 0.00 0.09 0.11
TM-FFA-ML 0.24 0.28 0.09 0.13 0.58 0.00 0.00 0.72 0.00 0.03 0.05 0.00 0.01 0.00 0.07 0.09 0.04 0.00 0.00 0.12 0.12
TM-FFA-WL 0.17 0.00 0.15 0.00 0.00 0.17 0.00 0.74 0.01 0.00 0.01 0.00 0.02 0.00 0.00 0.00 0.00 0.08 0.16 0.16 0.08
TM-FFA-CA 0.15 0.14 0.01 0.06 0.21 0.87 0.00 0.76 0.01 0.02 0.12 0.01 0.00 0.04 0.03 0.00 0.00 0.45 0.63 0.12 0.18
TM-FFA-IL 0.03 0.06 0.03 0.02 0.02 0.42 0.00 0.76 0.00 0.00 0.09 0.00 0.03 0.04 0.00 0.00 0.00 0.00 0.01 0.04 0.08
TM-FFA-IC 0.05 0.14 0.00 0.04 0.15 0.09 0.01 0.66 0.01 0.00 0.05 0.01 0.00 0.01 0.00 0.00 0.00 0.29 0.44 0.01 0.10
All 0.14 0.10 0.07 0.06 0.10 0.35 0.03 0.76 0.01 0.01 0.07 0.01 0.02 0.01 0.14 0.05 0.04 0.15 0.19 0.06 0.12
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 14/27
Best Setup: SM-FFA-ML
Best loop structures: ML and CA
(16% success rate)
![Page 127: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/127.jpg)
Setup po2 su2 po3 su3 fac gcd exp ℓ20 prm ssq sra srb ild lsb mul qad mod mi5 ma5 sm5 All
DIR 0.10 0.10 0.05 0.05 0.07 0.29 0.03 0.66 0.00 0.02 0.06 0.00 0.03 0.01 0.12 0.01 0.03 0.10 0.13 0.06 0.10
FFA 0.18 0.11 0.09 0.07 0.13 0.41 0.03 0.86 0.01 0.01 0.08 0.02 0.02 0.01 0.17 0.09 0.05 0.20 0.24 0.07 0.14
SM 0.16 0.12 0.09 0.07 0.07 0.48 0.05 0.87 0.01 0.02 0.08 0.02 0.03 0.01 0.23 0.07 0.07 0.20 0.23 0.04 0.15
TM 0.12 0.09 0.05 0.05 0.13 0.22 0.00 0.65 0.00 0.01 0.06 0.00 0.02 0.01 0.06 0.04 0.01 0.09 0.14 0.09 0.09
SM-DIR-CL 0.14 0.03 0.04 0.01 0.01 0.37 0.20 0.93 0.00 0.00 0.01 0.00 0.00 0.00 0.44 0.07 0.22 0.04 0.07 0.11 0.13
SM-DIR-ML 0.18 0.66 0.08 0.32 0.27 0.36 0.11 0.70 0.00 0.21 0.03 0.00 0.01 0.00 0.63 0.01 0.13 0.00 0.00 0.04 0.19
SM-DIR-WL 0.11 0.00 0.08 0.00 0.00 0.42 0.00 0.51 0.00 0.00 0.01 0.00 0.00 0.01 0.00 0.00 0.00 0.00 0.00 0.07 0.06
SM-DIR-CA 0.11 0.00 0.08 0.00 0.00 0.55 0.00 0.93 0.01 0.00 0.14 0.01 0.13 0.03 0.00 0.00 0.00 0.40 0.50 0.00 0.14
SM-DIR-IL 0.09 0.01 0.11 0.01 0.00 0.10 0.00 0.81 0.00 0.00 0.04 0.00 0.01 0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.06
SM-DIR-IC 0.04 0.03 0.00 0.00 0.00 0.68 0.00 0.88 0.01 0.00 0.13 0.02 0.05 0.02 0.01 0.00 0.00 0.43 0.55 0.02 0.14
SM-FFA-CL 0.32 0.03 0.16 0.00 0.03 0.41 0.19 1.00 0.00 0.00 0.01 0.07 0.00 0.00 0.72 0.41 0.31 0.19 0.17 0.05 0.20
SM-FFA-ML 0.14 0.57 0.15 0.46 0.45 0.47 0.12 0.92 0.02 0.05 0.05 0.00 0.00 0.00 0.89 0.30 0.15 0.00 0.00 0.06 0.24
SM-FFA-WL 0.32 0.00 0.17 0.02 0.04 0.46 0.00 0.90 0.01 0.00 0.01 0.01 0.05 0.01 0.00 0.00 0.00 0.00 0.02 0.06 0.10
SM-FFA-CA 0.17 0.00 0.05 0.00 0.02 0.80 0.00 0.97 0.02 0.00 0.22 0.06 0.02 0.01 0.00 0.00 0.00 0.64 0.74 0.00 0.19
SM-FFA-IL 0.21 0.03 0.11 0.00 0.00 0.42 0.01 0.93 0.01 0.00 0.15 0.01 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.05 0.10
SM-FFA-IC 0.12 0.02 0.01 0.03 0.00 0.76 0.00 0.90 0.05 0.00 0.16 0.03 0.01 0.01 0.01 0.00 0.00 0.72 0.74 0.04 0.18
TM-DIR-CL 0.12 0.07 0.07 0.03 0.01 0.00 0.00 0.81 0.00 0.00 0.00 0.00 0.00 0.00 0.26 0.05 0.04 0.00 0.00 0.13 0.08
TM-DIR-ML 0.17 0.27 0.04 0.14 0.43 0.01 0.00 0.30 0.00 0.01 0.18 0.00 0.03 0.00 0.03 0.00 0.01 0.00 0.00 0.14 0.09
TM-DIR-WL 0.15 0.00 0.07 0.00 0.00 0.11 0.00 0.24 0.00 0.00 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.01 0.03 0.08 0.04
TM-DIR-CA 0.04 0.05 0.04 0.08 0.05 0.83 0.00 0.66 0.00 0.00 0.11 0.01 0.06 0.01 0.04 0.01 0.00 0.24 0.30 0.10 0.13
TM-DIR-IL 0.03 0.01 0.02 0.00 0.00 0.04 0.00 0.75 0.00 0.00 0.02 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.01 0.00 0.05
TM-DIR-IC 0.04 0.03 0.00 0.01 0.09 0.05 0.00 0.45 0.00 0.00 0.04 0.00 0.03 0.02 0.00 0.00 0.00 0.05 0.15 0.06 0.05
TM-FFA-CL 0.20 0.04 0.12 0.06 0.01 0.00 0.02 1.00 0.00 0.00 0.00 0.00 0.01 0.00 0.34 0.32 0.07 0.00 0.00 0.09 0.11
TM-FFA-ML 0.24 0.28 0.09 0.13 0.58 0.00 0.00 0.72 0.00 0.03 0.05 0.00 0.01 0.00 0.07 0.09 0.04 0.00 0.00 0.12 0.12
TM-FFA-WL 0.17 0.00 0.15 0.00 0.00 0.17 0.00 0.74 0.01 0.00 0.01 0.00 0.02 0.00 0.00 0.00 0.00 0.08 0.16 0.16 0.08
TM-FFA-CA 0.15 0.14 0.01 0.06 0.21 0.87 0.00 0.76 0.01 0.02 0.12 0.01 0.00 0.04 0.03 0.00 0.00 0.45 0.63 0.12 0.18
TM-FFA-IL 0.03 0.06 0.03 0.02 0.02 0.42 0.00 0.76 0.00 0.00 0.09 0.00 0.03 0.04 0.00 0.00 0.00 0.00 0.01 0.04 0.08
TM-FFA-IC 0.05 0.14 0.00 0.04 0.15 0.09 0.01 0.66 0.01 0.00 0.05 0.01 0.00 0.01 0.00 0.00 0.00 0.29 0.44 0.01 0.10
All 0.14 0.10 0.07 0.06 0.10 0.35 0.03 0.76 0.01 0.01 0.07 0.01 0.02 0.01 0.14 0.05 0.04 0.15 0.19 0.06 0.12
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 14/27
Best Setup: SM-FFA-ML
Best loop structures: ML and CA
(16% success rate)
Worst loop structures: IL and WL
![Page 128: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/128.jpg)
Setup po2 su2 po3 su3 fac gcd exp ℓ20 prm ssq sra srb ild lsb mul qad mod mi5 ma5 sm5 All
DIR 0.10 0.10 0.05 0.05 0.07 0.29 0.03 0.66 0.00 0.02 0.06 0.00 0.03 0.01 0.12 0.01 0.03 0.10 0.13 0.06 0.10
FFA 0.18 0.11 0.09 0.07 0.13 0.41 0.03 0.86 0.01 0.01 0.08 0.02 0.02 0.01 0.17 0.09 0.05 0.20 0.24 0.07 0.14
SM 0.16 0.12 0.09 0.07 0.07 0.48 0.05 0.87 0.01 0.02 0.08 0.02 0.03 0.01 0.23 0.07 0.07 0.20 0.23 0.04 0.15
TM 0.12 0.09 0.05 0.05 0.13 0.22 0.00 0.65 0.00 0.01 0.06 0.00 0.02 0.01 0.06 0.04 0.01 0.09 0.14 0.09 0.09
SM-DIR-CL 0.14 0.03 0.04 0.01 0.01 0.37 0.20 0.93 0.00 0.00 0.01 0.00 0.00 0.00 0.44 0.07 0.22 0.04 0.07 0.11 0.13
SM-DIR-ML 0.18 0.66 0.08 0.32 0.27 0.36 0.11 0.70 0.00 0.21 0.03 0.00 0.01 0.00 0.63 0.01 0.13 0.00 0.00 0.04 0.19
SM-DIR-WL 0.11 0.00 0.08 0.00 0.00 0.42 0.00 0.51 0.00 0.00 0.01 0.00 0.00 0.01 0.00 0.00 0.00 0.00 0.00 0.07 0.06
SM-DIR-CA 0.11 0.00 0.08 0.00 0.00 0.55 0.00 0.93 0.01 0.00 0.14 0.01 0.13 0.03 0.00 0.00 0.00 0.40 0.50 0.00 0.14
SM-DIR-IL 0.09 0.01 0.11 0.01 0.00 0.10 0.00 0.81 0.00 0.00 0.04 0.00 0.01 0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.06
SM-DIR-IC 0.04 0.03 0.00 0.00 0.00 0.68 0.00 0.88 0.01 0.00 0.13 0.02 0.05 0.02 0.01 0.00 0.00 0.43 0.55 0.02 0.14
SM-FFA-CL 0.32 0.03 0.16 0.00 0.03 0.41 0.19 1.00 0.00 0.00 0.01 0.07 0.00 0.00 0.72 0.41 0.31 0.19 0.17 0.05 0.20
SM-FFA-ML 0.14 0.57 0.15 0.46 0.45 0.47 0.12 0.92 0.02 0.05 0.05 0.00 0.00 0.00 0.89 0.30 0.15 0.00 0.00 0.06 0.24
SM-FFA-WL 0.32 0.00 0.17 0.02 0.04 0.46 0.00 0.90 0.01 0.00 0.01 0.01 0.05 0.01 0.00 0.00 0.00 0.00 0.02 0.06 0.10
SM-FFA-CA 0.17 0.00 0.05 0.00 0.02 0.80 0.00 0.97 0.02 0.00 0.22 0.06 0.02 0.01 0.00 0.00 0.00 0.64 0.74 0.00 0.19
SM-FFA-IL 0.21 0.03 0.11 0.00 0.00 0.42 0.01 0.93 0.01 0.00 0.15 0.01 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.05 0.10
SM-FFA-IC 0.12 0.02 0.01 0.03 0.00 0.76 0.00 0.90 0.05 0.00 0.16 0.03 0.01 0.01 0.01 0.00 0.00 0.72 0.74 0.04 0.18
TM-DIR-CL 0.12 0.07 0.07 0.03 0.01 0.00 0.00 0.81 0.00 0.00 0.00 0.00 0.00 0.00 0.26 0.05 0.04 0.00 0.00 0.13 0.08
TM-DIR-ML 0.17 0.27 0.04 0.14 0.43 0.01 0.00 0.30 0.00 0.01 0.18 0.00 0.03 0.00 0.03 0.00 0.01 0.00 0.00 0.14 0.09
TM-DIR-WL 0.15 0.00 0.07 0.00 0.00 0.11 0.00 0.24 0.00 0.00 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.01 0.03 0.08 0.04
TM-DIR-CA 0.04 0.05 0.04 0.08 0.05 0.83 0.00 0.66 0.00 0.00 0.11 0.01 0.06 0.01 0.04 0.01 0.00 0.24 0.30 0.10 0.13
TM-DIR-IL 0.03 0.01 0.02 0.00 0.00 0.04 0.00 0.75 0.00 0.00 0.02 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.01 0.00 0.05
TM-DIR-IC 0.04 0.03 0.00 0.01 0.09 0.05 0.00 0.45 0.00 0.00 0.04 0.00 0.03 0.02 0.00 0.00 0.00 0.05 0.15 0.06 0.05
TM-FFA-CL 0.20 0.04 0.12 0.06 0.01 0.00 0.02 1.00 0.00 0.00 0.00 0.00 0.01 0.00 0.34 0.32 0.07 0.00 0.00 0.09 0.11
TM-FFA-ML 0.24 0.28 0.09 0.13 0.58 0.00 0.00 0.72 0.00 0.03 0.05 0.00 0.01 0.00 0.07 0.09 0.04 0.00 0.00 0.12 0.12
TM-FFA-WL 0.17 0.00 0.15 0.00 0.00 0.17 0.00 0.74 0.01 0.00 0.01 0.00 0.02 0.00 0.00 0.00 0.00 0.08 0.16 0.16 0.08
TM-FFA-CA 0.15 0.14 0.01 0.06 0.21 0.87 0.00 0.76 0.01 0.02 0.12 0.01 0.00 0.04 0.03 0.00 0.00 0.45 0.63 0.12 0.18
TM-FFA-IL 0.03 0.06 0.03 0.02 0.02 0.42 0.00 0.76 0.00 0.00 0.09 0.00 0.03 0.04 0.00 0.00 0.00 0.00 0.01 0.04 0.08
TM-FFA-IC 0.05 0.14 0.00 0.04 0.15 0.09 0.01 0.66 0.01 0.00 0.05 0.01 0.00 0.01 0.00 0.00 0.00 0.29 0.44 0.01 0.10
All 0.14 0.10 0.07 0.06 0.10 0.35 0.03 0.76 0.01 0.01 0.07 0.01 0.02 0.01 0.14 0.05 0.04 0.15 0.19 0.06 0.12
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 14/27
Easiest problems: ℓ20, gcd, ma5
![Page 129: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/129.jpg)
Setup po2 su2 po3 su3 fac gcd exp ℓ20 prm ssq sra srb ild lsb mul qad mod mi5 ma5 sm5 All
DIR 0.10 0.10 0.05 0.05 0.07 0.29 0.03 0.66 0.00 0.02 0.06 0.00 0.03 0.01 0.12 0.01 0.03 0.10 0.13 0.06 0.10
FFA 0.18 0.11 0.09 0.07 0.13 0.41 0.03 0.86 0.01 0.01 0.08 0.02 0.02 0.01 0.17 0.09 0.05 0.20 0.24 0.07 0.14
SM 0.16 0.12 0.09 0.07 0.07 0.48 0.05 0.87 0.01 0.02 0.08 0.02 0.03 0.01 0.23 0.07 0.07 0.20 0.23 0.04 0.15
TM 0.12 0.09 0.05 0.05 0.13 0.22 0.00 0.65 0.00 0.01 0.06 0.00 0.02 0.01 0.06 0.04 0.01 0.09 0.14 0.09 0.09
SM-DIR-CL 0.14 0.03 0.04 0.01 0.01 0.37 0.20 0.93 0.00 0.00 0.01 0.00 0.00 0.00 0.44 0.07 0.22 0.04 0.07 0.11 0.13
SM-DIR-ML 0.18 0.66 0.08 0.32 0.27 0.36 0.11 0.70 0.00 0.21 0.03 0.00 0.01 0.00 0.63 0.01 0.13 0.00 0.00 0.04 0.19
SM-DIR-WL 0.11 0.00 0.08 0.00 0.00 0.42 0.00 0.51 0.00 0.00 0.01 0.00 0.00 0.01 0.00 0.00 0.00 0.00 0.00 0.07 0.06
SM-DIR-CA 0.11 0.00 0.08 0.00 0.00 0.55 0.00 0.93 0.01 0.00 0.14 0.01 0.13 0.03 0.00 0.00 0.00 0.40 0.50 0.00 0.14
SM-DIR-IL 0.09 0.01 0.11 0.01 0.00 0.10 0.00 0.81 0.00 0.00 0.04 0.00 0.01 0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.06
SM-DIR-IC 0.04 0.03 0.00 0.00 0.00 0.68 0.00 0.88 0.01 0.00 0.13 0.02 0.05 0.02 0.01 0.00 0.00 0.43 0.55 0.02 0.14
SM-FFA-CL 0.32 0.03 0.16 0.00 0.03 0.41 0.19 1.00 0.00 0.00 0.01 0.07 0.00 0.00 0.72 0.41 0.31 0.19 0.17 0.05 0.20
SM-FFA-ML 0.14 0.57 0.15 0.46 0.45 0.47 0.12 0.92 0.02 0.05 0.05 0.00 0.00 0.00 0.89 0.30 0.15 0.00 0.00 0.06 0.24
SM-FFA-WL 0.32 0.00 0.17 0.02 0.04 0.46 0.00 0.90 0.01 0.00 0.01 0.01 0.05 0.01 0.00 0.00 0.00 0.00 0.02 0.06 0.10
SM-FFA-CA 0.17 0.00 0.05 0.00 0.02 0.80 0.00 0.97 0.02 0.00 0.22 0.06 0.02 0.01 0.00 0.00 0.00 0.64 0.74 0.00 0.19
SM-FFA-IL 0.21 0.03 0.11 0.00 0.00 0.42 0.01 0.93 0.01 0.00 0.15 0.01 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.05 0.10
SM-FFA-IC 0.12 0.02 0.01 0.03 0.00 0.76 0.00 0.90 0.05 0.00 0.16 0.03 0.01 0.01 0.01 0.00 0.00 0.72 0.74 0.04 0.18
TM-DIR-CL 0.12 0.07 0.07 0.03 0.01 0.00 0.00 0.81 0.00 0.00 0.00 0.00 0.00 0.00 0.26 0.05 0.04 0.00 0.00 0.13 0.08
TM-DIR-ML 0.17 0.27 0.04 0.14 0.43 0.01 0.00 0.30 0.00 0.01 0.18 0.00 0.03 0.00 0.03 0.00 0.01 0.00 0.00 0.14 0.09
TM-DIR-WL 0.15 0.00 0.07 0.00 0.00 0.11 0.00 0.24 0.00 0.00 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.01 0.03 0.08 0.04
TM-DIR-CA 0.04 0.05 0.04 0.08 0.05 0.83 0.00 0.66 0.00 0.00 0.11 0.01 0.06 0.01 0.04 0.01 0.00 0.24 0.30 0.10 0.13
TM-DIR-IL 0.03 0.01 0.02 0.00 0.00 0.04 0.00 0.75 0.00 0.00 0.02 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.01 0.00 0.05
TM-DIR-IC 0.04 0.03 0.00 0.01 0.09 0.05 0.00 0.45 0.00 0.00 0.04 0.00 0.03 0.02 0.00 0.00 0.00 0.05 0.15 0.06 0.05
TM-FFA-CL 0.20 0.04 0.12 0.06 0.01 0.00 0.02 1.00 0.00 0.00 0.00 0.00 0.01 0.00 0.34 0.32 0.07 0.00 0.00 0.09 0.11
TM-FFA-ML 0.24 0.28 0.09 0.13 0.58 0.00 0.00 0.72 0.00 0.03 0.05 0.00 0.01 0.00 0.07 0.09 0.04 0.00 0.00 0.12 0.12
TM-FFA-WL 0.17 0.00 0.15 0.00 0.00 0.17 0.00 0.74 0.01 0.00 0.01 0.00 0.02 0.00 0.00 0.00 0.00 0.08 0.16 0.16 0.08
TM-FFA-CA 0.15 0.14 0.01 0.06 0.21 0.87 0.00 0.76 0.01 0.02 0.12 0.01 0.00 0.04 0.03 0.00 0.00 0.45 0.63 0.12 0.18
TM-FFA-IL 0.03 0.06 0.03 0.02 0.02 0.42 0.00 0.76 0.00 0.00 0.09 0.00 0.03 0.04 0.00 0.00 0.00 0.00 0.01 0.04 0.08
TM-FFA-IC 0.05 0.14 0.00 0.04 0.15 0.09 0.01 0.66 0.01 0.00 0.05 0.01 0.00 0.01 0.00 0.00 0.00 0.29 0.44 0.01 0.10
All 0.14 0.10 0.07 0.06 0.10 0.35 0.03 0.76 0.01 0.01 0.07 0.01 0.02 0.01 0.14 0.05 0.04 0.15 0.19 0.06 0.12
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 14/27
Easiest problems: ℓ20, gcd, ma5
Hardest problems: prm, lsb, srb, ssq
![Page 130: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/130.jpg)
1 Benchmark Problems
2 Investigated Ideas
3 Experiments
4 Summary
5 Evolved Programs
SummaryEvolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 15/27
![Page 131: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/131.jpg)
Summary
Evolution of non-trivial exact integer algorithms is possible
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 16/27
![Page 132: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/132.jpg)
Summary
Evolution of non-trivial exact integer algorithms is possible
. . . but a hard problem
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 16/27
![Page 133: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/133.jpg)
Summary
Evolution of non-trivial exact integer algorithms is possible
. . . but a hard problem
Search for novel behaviors [6, 7] (via FFA [5]) = good
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 16/27
![Page 134: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/134.jpg)
Summary
Evolution of non-trivial exact integer algorithms is possible
. . . but a hard problem
Search for novel behaviors [6, 7] (via FFA [5]) = good
Transactional memory idea = not good
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 16/27
![Page 135: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/135.jpg)
Summary
Evolution of non-trivial exact integer algorithms is possible
. . . but a hard problem
Search for novel behaviors [6, 7] (via FFA [5]) = good
Transactional memory idea = not good
Benchmark set: will hopefully be interesting for a long time
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 16/27
![Page 136: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/136.jpg)
1 Benchmark Problems
2 Investigated Ideas
3 Experiments
4 Summary
5 Evolved Programs
Evolved ProgramsEvolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 17/27
![Page 137: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/137.jpg)
CL-SM program for srb
Computes square root of a number
Contains the same code twice . . . remove one copy and it stops
working for the single input 5
Listing 1: Manual Translation to Java
static int srb(int m0) {
int m2 = 0, t = 0;
for (int i = m0; i > 0; i--) {
t = m0 / (1 + m2);
m2 = (1 + m2 - ((t != 0) , (m2 / t) , m2));
t = m0 / (1 + m2);
m2 = (1 + m2 - ((t != 0) , (m2 / t) , m2));
}
return m2;
}
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 18/27
![Page 138: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/138.jpg)
CA-TM program for lsb
Listing 2: Manual Translation to Java
static int gpLSB(int m0) {
int m1 = 0, m2 = 0, tm0 = m0, tm2 = m2;
for (;;) { // small m0 -> many iterations
if (m0 != 0) {
tm2 = (m1 - m0);
m1 = 1;
if (m2 != 0) {
tm2 = (m2 - m0);
if (tm2 != 0) tm0 = (m0 % tm2);
}
}
if ((tm0 == m0) && (tm2 == m2))
return m2;
m0 = tm0;
m2 = tm2;
}
}
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 19/27
Calculates value of least
significant bit
Faster for larger inputs (probably
utilizes integer overflow. . . )
![Page 139: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/139.jpg)
CL-TM program for qad
Calculates (m0 − 1)(m0 − 2) without multiplication
Listing 3: Manual Translation to Java
static int qad(int m0) {
int m1 = 0, t = (m0 - 1), tm1 = 0;
t = m0 - 1;
for (int i = t; i > 0; i--) {
tm1 = (m1 + m0);
t = m0 + m0 + m1 - 1;
m1 = tm1;
}
return (m0 - (1 - t));
}
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 20/27
![Page 140: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/140.jpg)
CA-TM program for gcd
Listing 4: Manual Translation to Java
static int gcd(int m0, int m1) {
int m2 = 0, tm0 = 0, tm1 = 0, tm2 = 0;
for (;;) {
if (m1 != 0)
tm1 = m0;
tm0 = (m0 != 0) , (m1 % m0) , m1;
if ((m0 != 0) && (m1 != 0) && ((m1 % m0) == 0))
tm2 = m0;
if ((m0 == tm0) && (m1 == tm1) && (m2 == tm2))
return m2;
m0 = tm0;
m1 = tm1;
m2 = tm2;
}
}
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 21/27
Calculates value of greatest
common divisor
![Page 141: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/141.jpg)
IC-SM program for mi5
Listing 5: Manual Translation to Java
static int mi5(int m0, int m1, int m2, int m3, int m4) {
int t;
for(;;) {
t = m4;
if(m4 > m0) m4 = m0;
if(m4 > m2) m4 = m2;
if((m4 == t) && (m2 == m1) &&
(m1 == m3)) return m4;
m2 = m1;
m1 = m3;
}
}
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 22/27
Calculates minimum of five
variables in a “serial” fashion
![Page 142: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/142.jpg)
WL-SM program for l20
Checks if a number is less than 20, without having any constant
larger than 1
(the constants in the listing are simplified results of calculations)
Listing 6: Manual Translation to Java
static int l20(int m0) {
int t = ((m0 - 2) / 3) / 3; // integer division!
if (t < 2) return 1;
else return 0;
}
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 23/27
![Page 143: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/143.jpg)
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 24/27
![Page 144: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/144.jpg)
ReferencesEvolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 25/27
![Page 145: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/145.jpg)
References I
1. Mingxu Wan, Thomas Weise, and Ke Tang. Novel loop structures and the evolution of mathematical algorithms. In Sara Silva,
James A. Foster, Miguel Nicolau, Penousal Machado, and Mario Giacobini, editors, Proceedings of the 14th European Conference
on Genetic Programming (EuroGP’11), volume 6621/2011 of Lecture Notes in Computer Science (LNCS), pages 49–60, Torino,
Italy, April 27–29, 2011. Berlin, Germany: Springer-Verlag GmbH. 10.1007/978-3-642-20407-4_5. URL
http://www.it-weise.de/documents/files/WWT2011NLSATEOMA.pdf.
2. Guang Chen and Mengjie Zhang. Evolving while-loop structures in genetic programming for factorial and ant problems. In Shichao
Zhang and Ray Jarvis, editors, Advances in Artificial Intelligence. Proceedings of the 18th Australian Joint Conference on Artificial
Intelligence (AI’05), volume 3809/2005 of Lecture Notes in Computer Science (LNCS), pages 1079–1085, Sydney, NSW, Australia:
University of Technology, Sydney (UTS), December 5–9, 2005. Berlin, Germany: Springer-Verlag GmbH. 10.1007/11589990_144.
3. Timothy Lai. Discovery of understandable math formulas using genetic programming. In Genetic Algorithms and Genetic
Programming at Stanford, pages 118–127. Stanford, CA, USA: Stanford University Bookstore, Stanford University, Fall 2003. URL
http://www.genetic-programming.org/sp2003/Lai.pdf.
4. Thomas Weise and Ke Tang. Evolving distributed algorithms with genetic programming. IEEE Transactions on Evolutionary
Computation (IEEE-EC), 16(2):242–265, April 2012. 10.1109/TEVC.2011.2112666. URL
http://www.it-weise.de/documents/files/WT2011EVDAWGP.pdf.
5. Thomas Weise, Mingxu Wan, Ke Tang, Pu Wang, Alexandre Devert, and Xin Yao. Frequency fitness assignment. IEEE Transactions
on Evolutionary Computation (IEEE-EC), 18(2):226–243, April 2014. 10.1109/TEVC.2013.2251885. URL
http://www.it-weise.de/documents/files/WWTWDY2014FFA.pdf.
6. Joel Lehman and Kenneth Owen Stanley. Abandoning objectives: Evolution through the search for novelty alone. Evolutionary
Computation, 19(2):189–223, Summer 2011. 10.1162/EVCO_a_00025. URL
http://eplex.cs.ucf.edu/papers/lehman_ecj10.pdf.
7. Marcus Hutter and Shane Legg. Fitness uniform optimization. IEEE Transactions on Evolutionary Computation (IEEE-EC), 10(5):
568–589, October 2006. 10.1109/TEVC.2005.863127. URL http://arxiv.org/abs/cs/0610126v1.
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 26/27
![Page 146: Evolving Exact Algorithms with GP - hfuu.edu.cn](https://reader031.vdocuments.us/reader031/viewer/2022020621/61e90e288da15164304cb8d3/html5/thumbnails/146.jpg)
Citation Information
@inproceedings{WWTY2014EEIAWGP,
author = {Thomas Weise and Mingxu Wan and Ke Tang and Xin Yao},
title = {{Evolving Exact Integer Algorithms with Genetic
Programming}},
booktitle = {Proceedings of the IEEE Congress on Evolutionary
Computation (CEC’14), Proceedings of the 2014 World
Congress on Computational Intelligence (WCCI’14)},
publisher = {Los Alamitos, CA, USA: IEEE Computer Society Press},
address = {Beijing, China: Beijing International Convention
Center (BICC)},
pages = {1816--1823},
year = {2014},
month = jul # {~6--11, },
doi = {10.1109/CEC.2014.6900292},
isbn = {978-1-4799-1488-3},
},
Evolving Exact Algorithms with GP, CEC @ WCCI, 2014-07-10, Beijing Thomas Weise · iao.hfuu.edu.cn · [email protected] 27/27