introduction to arnoldi method - matematik | kth · main eigenvalue algorithms in this course...
TRANSCRIPT
![Page 1: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/1.jpg)
Introduction to Arnoldi methodSF2524 - Matrix Computations for Large-scale Systems
KTH Royal Institute of Technology (Elias Jarlebring)
2014-11-07
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 1 / 9
![Page 2: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/2.jpg)
Agenda lecture 2
Introduction to Arnoldi method
Gram-Schmidt - efficiency and roundoff errors
Derivation of Arnoldi method
(Convergence characterization)
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 2 / 9
![Page 3: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/3.jpg)
Main eigenvalue algorithms in this course
Fundamental eigenvalue techniques (Lecture 1)
Arnoldi method (Lecture 2-3).
Typically suitable whenI we are interested in a small number of eigenvalues,I the matrix is large and sparseI Currently solvable size on desktop m ∼ 106 (depending on structure)
QR-method (Lecture 8-9).
Typically suitable whenI we want to compute all eigenvalues,I the matrix does not have any particular easy structure.I Solvable size on desktop m ∼ 1000.
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 3 / 9
![Page 4: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/4.jpg)
Main eigenvalue algorithms in this course
Fundamental eigenvalue techniques (Lecture 1)
Arnoldi method (Lecture 2-3).Typically suitable when
I we are interested in a small number of eigenvalues,I the matrix is large and sparseI Currently solvable size on desktop m ∼ 106 (depending on structure)
QR-method (Lecture 8-9).Typically suitable when
I we want to compute all eigenvalues,I the matrix does not have any particular easy structure.I Solvable size on desktop m ∼ 1000.
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 3 / 9
![Page 5: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/5.jpg)
Finite-element method example (slide 1/2)Laplace eigenvalue problem
∆u = λu
with Dirichlet boundary conditions outside wing profile.
0 1 2 3 4−2
−1
0
1
2
0
2
4
6
8
10
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 4 / 9
![Page 6: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/6.jpg)
Finite-element method example (slide 2/2)
⇒ Problem: Compute eigenvalues of matrix A ∈ Rm×m where m large butfew non-zero elements:
0 200 400
0
100
200
300
400
500
nz = 3550
In MATLAB with very efficient QR-method:tic; eig(full(A)); toc
More than 10 hours ⇒ we need a different method.
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 5 / 9
![Page 7: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/7.jpg)
Finite-element method example (slide 2/2)
⇒ Problem: Compute eigenvalues of matrix A ∈ Rm×m where m large butfew non-zero elements:
0 200 400
0
100
200
300
400
500
nz = 3550
In MATLAB with very efficient QR-method:tic; eig(full(A)); toc
More than 10 hours ⇒ we need a different method.
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 5 / 9
![Page 8: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/8.jpg)
Finite-element method example (slide 2/2)
⇒ Problem: Compute eigenvalues of matrix A ∈ Rm×m where m large butfew non-zero elements:
0 200 400
0
100
200
300
400
500
nz = 3550
In MATLAB with very efficient QR-method:tic; eig(full(A)); toc
More than 10 hours ⇒ we need a different method.
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 5 / 9
![Page 9: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/9.jpg)
Idea of Arnoldi method (slide 1/3)
Eigenvalue problem
Ax = λx .
Suppose Q = (q1, . . . , qn) ∈ Rm×n and
x ∈ span(q1, . . . , qn)
⇒ There exists z ∈ Rn such that
AQz = λQz
and
Q orthogonal
QTAQz = λQTQz
= λz
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 6 / 9
![Page 10: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/10.jpg)
Idea of Arnoldi method (slide 1/3)
Eigenvalue problem
Ax = λx .
Suppose Q = (q1, . . . , qn) ∈ Rm×n and
x ∈ span(q1, . . . , qn)
⇒ There exists z ∈ Rn such that
AQz = λQz
and
Q orthogonal
QTAQz = λQTQz
= λz
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 6 / 9
![Page 11: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/11.jpg)
Idea of Arnoldi method (slide 1/3)
Eigenvalue problem
Ax = λx .
Suppose Q = (q1, . . . , qn) ∈ Rm×n and
x ∈ span(q1, . . . , qn)
⇒ There exists z ∈ Rn such that
AQz = λQz
and
Q orthogonal
QTAQz = λQTQz
= λz
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 6 / 9
![Page 12: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/12.jpg)
Idea of Arnoldi method (slide 1/3)
Eigenvalue problem
Ax = λx .
Suppose Q = (q1, . . . , qn) ∈ Rm×n and
x ∈ span(q1, . . . , qn)
⇒ There exists z ∈ Rn such that
AQz = λQz
and
Q orthogonal
QTAQz = λQTQz
= λz
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 6 / 9
![Page 13: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/13.jpg)
Idea of Arnoldi method (slide 1/3)
Eigenvalue problem
Ax = λx .
Suppose Q = (q1, . . . , qn) ∈ Rm×n and
x ∈ span(q1, . . . , qn)
⇒ There exists z ∈ Rn such that
AQz = λQz
and Q orthogonalQTAQz = λQTQz = λz
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 6 / 9
![Page 14: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/14.jpg)
Idea of Arnoldi method (slide 2/3)
Rayleigh-Ritz procedure
Let Q ∈ Rm×n be orthogonal basis of some subspace. Solutions (µ, z) tothe eigenvalue problem
QTAQz = µz
are called Ritz pairs. The procedure (Rayleigh-Ritz)
returns an exact eigenvalue if x ∈ span q1, . . . , qn, and
approximates eigenvalues of A “well” if
x ≈ x̃ ∈ span(q1, . . . , qn).
What is a good subspace span(Q)?
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 7 / 9
![Page 15: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/15.jpg)
Idea of Arnoldi method (slide 2/3)
Rayleigh-Ritz procedure
Let Q ∈ Rm×n be orthogonal basis of some subspace.
Solutions (µ, z) tothe eigenvalue problem
QTAQz = µz
are called Ritz pairs. The procedure (Rayleigh-Ritz)
returns an exact eigenvalue if x ∈ span q1, . . . , qn, and
approximates eigenvalues of A “well” if
x ≈ x̃ ∈ span(q1, . . . , qn).
What is a good subspace span(Q)?
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 7 / 9
![Page 16: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/16.jpg)
Idea of Arnoldi method (slide 2/3)
Rayleigh-Ritz procedure
Let Q ∈ Rm×n be orthogonal basis of some subspace. Solutions (µ, z) tothe eigenvalue problem
QTAQz = µz
are called Ritz pairs.
The procedure (Rayleigh-Ritz)
returns an exact eigenvalue if x ∈ span q1, . . . , qn, and
approximates eigenvalues of A “well” if
x ≈ x̃ ∈ span(q1, . . . , qn).
What is a good subspace span(Q)?
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 7 / 9
![Page 17: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/17.jpg)
Idea of Arnoldi method (slide 2/3)
Rayleigh-Ritz procedure
Let Q ∈ Rm×n be orthogonal basis of some subspace. Solutions (µ, z) tothe eigenvalue problem
QTAQz = µz
are called Ritz pairs. The procedure (Rayleigh-Ritz)
returns an exact eigenvalue if x ∈ span q1, . . . , qn,
and
approximates eigenvalues of A “well” if
x ≈ x̃ ∈ span(q1, . . . , qn).
What is a good subspace span(Q)?
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 7 / 9
![Page 18: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/18.jpg)
Idea of Arnoldi method (slide 2/3)
Rayleigh-Ritz procedure
Let Q ∈ Rm×n be orthogonal basis of some subspace. Solutions (µ, z) tothe eigenvalue problem
QTAQz = µz
are called Ritz pairs. The procedure (Rayleigh-Ritz)
returns an exact eigenvalue if x ∈ span q1, . . . , qn, and
approximates eigenvalues of A “well” if
x ≈ x̃ ∈ span(q1, . . . , qn).
What is a good subspace span(Q)?
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 7 / 9
![Page 19: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/19.jpg)
Idea of Arnoldi method (slide 2/3)
Rayleigh-Ritz procedure
Let Q ∈ Rm×n be orthogonal basis of some subspace. Solutions (µ, z) tothe eigenvalue problem
QTAQz = µz
are called Ritz pairs. The procedure (Rayleigh-Ritz)
returns an exact eigenvalue if x ∈ span q1, . . . , qn, and
approximates eigenvalues of A “well” if
x ≈ x̃ ∈ span(q1, . . . , qn).
What is a good subspace span(Q)?
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 7 / 9
![Page 20: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/20.jpg)
Idea of Arnoldi method (slide 2/3)
Rayleigh-Ritz procedure
Let Q ∈ Rm×n be orthogonal basis of some subspace. Solutions (µ, z) tothe eigenvalue problem
QTAQz = µz
are called Ritz pairs. The procedure (Rayleigh-Ritz)
returns an exact eigenvalue if x ∈ span q1, . . . , qn, and
approximates eigenvalues of A “well” if
x ≈ x̃ ∈ span(q1, . . . , qn).
What is a good subspace span(Q)?
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 7 / 9
![Page 21: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/21.jpg)
Idea of Arnoldi method (slide 3/3)Recall: Power method approximates the largest eigenvalue well.
Definition: Krylov matrix / subspace
Kn(A, q1) := (q1,Aq1, . . .An−1q1)
Kn(A, q1) := span(q1,Aq1, . . .An−1q1).
Justification of Arnoldi method
Use Rayleigh-Ritz on Q = (q1, . . . , qn) where
span(q1, . . . , qn) = Kn(A, q1)
Arnoldi method is a “clever” procedure to construct Hn = V TAV .
“Clever”:
(1) We exploit sparsity (we will only need matrix-vector product)(2) We expand V with one row in each iteration⇒ Iterate until we are happy.
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 8 / 9
![Page 22: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/22.jpg)
Idea of Arnoldi method (slide 3/3)Recall: Power method approximates the largest eigenvalue well.
Definition: Krylov matrix / subspace
Kn(A, q1) := (q1,Aq1, . . .An−1q1)
Kn(A, q1) := span(q1,Aq1, . . .An−1q1).
Justification of Arnoldi method
Use Rayleigh-Ritz on Q = (q1, . . . , qn) where
span(q1, . . . , qn) = Kn(A, q1)
Arnoldi method is a “clever” procedure to construct Hn = V TAV .
“Clever”:
(1) We exploit sparsity (we will only need matrix-vector product)(2) We expand V with one row in each iteration⇒ Iterate until we are happy.
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 8 / 9
![Page 23: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/23.jpg)
Idea of Arnoldi method (slide 3/3)Recall: Power method approximates the largest eigenvalue well.
Definition: Krylov matrix / subspace
Kn(A, q1) := (q1,Aq1, . . .An−1q1)
Kn(A, q1) := span(q1,Aq1, . . .An−1q1).
Justification of Arnoldi method
Use Rayleigh-Ritz on Q = (q1, . . . , qn) where
span(q1, . . . , qn) = Kn(A, q1)
Arnoldi method is a “clever” procedure to construct Hn = V TAV .
“Clever”:
(1) We exploit sparsity (we will only need matrix-vector product)(2) We expand V with one row in each iteration⇒ Iterate until we are happy.
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 8 / 9
![Page 24: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/24.jpg)
Idea of Arnoldi method (slide 3/3)Recall: Power method approximates the largest eigenvalue well.
Definition: Krylov matrix / subspace
Kn(A, q1) := (q1,Aq1, . . .An−1q1)
Kn(A, q1) := span(q1,Aq1, . . .An−1q1).
Justification of Arnoldi method
Use Rayleigh-Ritz on Q = (q1, . . . , qn) where
span(q1, . . . , qn) = Kn(A, q1)
Arnoldi method is a “clever” procedure to construct Hn = V TAV .
“Clever”:
(1) We exploit sparsity (we will only need matrix-vector product)(2) We expand V with one row in each iteration⇒ Iterate until we are happy.
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 8 / 9
![Page 25: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/25.jpg)
Idea of Arnoldi method (slide 3/3)Recall: Power method approximates the largest eigenvalue well.
Definition: Krylov matrix / subspace
Kn(A, q1) := (q1,Aq1, . . .An−1q1)
Kn(A, q1) := span(q1,Aq1, . . .An−1q1).
Justification of Arnoldi method
Use Rayleigh-Ritz on Q = (q1, . . . , qn) where
span(q1, . . . , qn) = Kn(A, q1)
Arnoldi method is a “clever” procedure to construct Hn = V TAV .
“Clever”:
(1) We exploit sparsity (we will only need matrix-vector product)
(2) We expand V with one row in each iteration⇒ Iterate until we are happy.
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 8 / 9
![Page 26: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/26.jpg)
Idea of Arnoldi method (slide 3/3)Recall: Power method approximates the largest eigenvalue well.
Definition: Krylov matrix / subspace
Kn(A, q1) := (q1,Aq1, . . .An−1q1)
Kn(A, q1) := span(q1,Aq1, . . .An−1q1).
Justification of Arnoldi method
Use Rayleigh-Ritz on Q = (q1, . . . , qn) where
span(q1, . . . , qn) = Kn(A, q1)
Arnoldi method is a “clever” procedure to construct Hn = V TAV .
“Clever”:
(1) We exploit sparsity (we will only need matrix-vector product)(2) We expand V with one row in each iteration⇒ Iterate until we are happy.
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 8 / 9
![Page 27: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/27.jpg)
Idea of Arnoldi method (slide 3/3)Recall: Power method approximates the largest eigenvalue well.
Definition: Krylov matrix / subspace
Kn(A, q1) := (q1,Aq1, . . .An−1q1)
Kn(A, q1) := span(q1,Aq1, . . .An−1q1).
Justification of Arnoldi method
Use Rayleigh-Ritz on Q = (q1, . . . , qn) where
span(q1, . . . , qn) = Kn(A, q1)
Arnoldi method is a “clever” procedure to construct Hn = V TAV .
“Clever”:
(1) We exploit sparsity (we will only need matrix-vector product)(2) We expand V with one row in each iteration⇒ Iterate until we are happy.
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 8 / 9
![Page 28: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/28.jpg)
Arnoldi method graphicallyGraphical illustration of algorithm:
Q =q1 q2 q3 q4= H
w
A
w⊥
Orthogonalize
After iteration: Take eigenvalues of H as approximate eigenvalues.
* show arnoldi.m and Hessenberg matrix in matlab *
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 9 / 9
![Page 29: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/29.jpg)
Arnoldi method graphicallyGraphical illustration of algorithm:
Q =q1 q2 q3 q4= H
w
A
w⊥
Orthogonalize
After iteration: Take eigenvalues of H as approximate eigenvalues.
* show arnoldi.m and Hessenberg matrix in matlab *
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 9 / 9
![Page 30: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/30.jpg)
Arnoldi method graphicallyGraphical illustration of algorithm:
Q =q1 q2 q3 q4= H
w
A
w⊥
Orthogonalize
After iteration: Take eigenvalues of H as approximate eigenvalues.
* show arnoldi.m and Hessenberg matrix in matlab *
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 9 / 9
![Page 31: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/31.jpg)
Arnoldi method graphicallyGraphical illustration of algorithm:
Q =q1 q2 q3 q4= H
w
A
w⊥
Orthogonalize
After iteration: Take eigenvalues of H as approximate eigenvalues.
* show arnoldi.m and Hessenberg matrix in matlab *
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 9 / 9
![Page 32: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/32.jpg)
Arnoldi method graphicallyGraphical illustration of algorithm:
Q =q1 q2 q3 q4= H
w
A
w⊥
Orthogonalize
After iteration: Take eigenvalues of H as approximate eigenvalues.
* show arnoldi.m and Hessenberg matrix in matlab *
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 9 / 9
![Page 33: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/33.jpg)
Arnoldi method graphicallyGraphical illustration of algorithm:
Q =q1 q2 q3 q4= H
w
A
w⊥
Orthogonalize
After iteration: Take eigenvalues of H as approximate eigenvalues.
* show arnoldi.m and Hessenberg matrix in matlab *
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 9 / 9
![Page 34: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/34.jpg)
Arnoldi method graphicallyGraphical illustration of algorithm:
Q =q1 q2 q3 q4= H
w
A
w⊥
Orthogonalize
After iteration: Take eigenvalues of H as approximate eigenvalues.
* show arnoldi.m and Hessenberg matrix in matlab *
We will now...
(1) derive a good orthogonalization procedure: variants of Gram-Schmidt,
(2) show that Arnoldi generates a Rayleigh-Ritz approximation,
(3) characterize the convergence.
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 9 / 9
![Page 35: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/35.jpg)
Arnoldi method graphicallyGraphical illustration of algorithm:
Q =q1 q2 q3 q4= H
w
A
w⊥
Orthogonalize
After iteration: Take eigenvalues of H as approximate eigenvalues.
* show arnoldi.m and Hessenberg matrix in matlab *
We will now...
(1) derive a good orthogonalization procedure: variants of Gram-Schmidt,
(2) show that Arnoldi generates a Rayleigh-Ritz approximation,
(3) characterize the convergence.
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 9 / 9
![Page 36: Introduction to Arnoldi method - Matematik | KTH · Main eigenvalue algorithms in this course Fundamental eigenvalue techniques (Lecture 1) Arnoldi method (Lecture 2-3). Typically](https://reader034.vdocuments.us/reader034/viewer/2022050717/5e4b128d4333672ce320af89/html5/thumbnails/36.jpg)
Arnoldi method graphicallyGraphical illustration of algorithm:
Q =q1 q2 q3 q4= H
w
A
w⊥
Orthogonalize
After iteration: Take eigenvalues of H as approximate eigenvalues.
* show arnoldi.m and Hessenberg matrix in matlab *
We will now...
(1) derive a good orthogonalization procedure: variants of Gram-Schmidt,
(2) show that Arnoldi generates a Rayleigh-Ritz approximation,
(3) characterize the convergence.
KTH Royal Institute of Technology (Elias Jarlebring)Introduction to Arnoldi method 2014-11-07 9 / 9