spectral graph theory and its applicationscs- graph theory and its applications daniel a. spielman...
TRANSCRIPT
![Page 1: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/1.jpg)
Spectral Graph Theoryand its
Applications
Daniel A. SpielmanDept. of Computer Science
Program in Applied MathematicsYale Unviersity
![Page 2: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/2.jpg)
OutlineAdjacency matrix and Laplacian
Intuition, spectral graph drawing
Physical intuition
Isomorphism testing
Random walks
Graph Partitioning and clustering
Distributions of eigenvalues and compression
Computation
![Page 3: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/3.jpg)
What I’m Skipping
Matrix-tree theorem.
Most of algebraic graph theory.
Special graphs (e.g. Cayley graphs).
Connections to codes and designs.
Lots of work by theorists.
Expanders.
![Page 4: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/4.jpg)
The Adjacency Matrix1 2 3 4
λ is eigenvalue and v is eigenvector if
Think of , or even better
Symmetric -> n real eigenvalues andreal eigenvectors form orthonormal basis
![Page 5: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/5.jpg)
Example
1 2 3 4
-1 -0.618 0.618 1
![Page 6: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/6.jpg)
Example
1 2 3 4
-1 -0.618 0.618 1
1-0.6180.382
![Page 7: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/7.jpg)
Example: invariant under re-labeling
1 2 3 4
-1 -0.618 0.618 1
![Page 8: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/8.jpg)
Example: invariant under re-labeling
2 1 3 4
-1 -0.618 0.618 1
![Page 9: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/9.jpg)
Operators and Quadratic Forms
View of A as an operator:
View of A as quadratic form:
if and then
![Page 10: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/10.jpg)
Laplacian: natural quadratic form on graphs
where D is diagonal matrix of degrees
1 2 3 4
![Page 11: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/11.jpg)
Laplacian: fast facts
so, zero is an eigenvalue
If k connected components,
Fiedler (‘73) called “algebraic connectivity of a graph”
The further from 0, the more connected.
![Page 12: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/12.jpg)
Embedding graph in line (Hall ’70)
map
minimize
trivial solution: So, require
Solution
Atkins, Boman, Hendrickson ’97:Gives correct embedding for graphs like
![Page 13: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/13.jpg)
Courant-Fischer definition of eigvals/vecs
(here )
![Page 14: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/14.jpg)
Embedding graph in plane (Hall ’70)
minimize
map
trivial solution:
Also require
Solution up to rotation
So, require
degenerate solution:
![Page 15: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/15.jpg)
A Graph
![Page 16: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/16.jpg)
Drawing of the graph using v2, v3
Plot vertex i at
![Page 17: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/17.jpg)
![Page 18: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/18.jpg)
Spectral drawing of Streets in Rome
![Page 19: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/19.jpg)
Spectral drawing of Erdos graph:edge between co-authors of papers
![Page 20: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/20.jpg)
Dodecahedron
Best embedded by first three eigenvectors
![Page 21: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/21.jpg)
Condition for eigenvector
Spectral graph drawing: Tutte justification
Gives for all i
λ small says x(i) near average of neighbors
Tutte ‘63: If fix outside face, and let everyother vertex be average of neighbors, getplanar embedding of planar graph.
![Page 22: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/22.jpg)
Tutte ‘63 embedding of a graph.
Fix outside face.Edges -> springs.
Vertex at centerof mass of nbrs.
3-connected -> get planar embedding
![Page 23: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/23.jpg)
Fundamental modes: string with fixed ends
![Page 24: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/24.jpg)
Fundamental modes: string with free ends
![Page 25: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/25.jpg)
Eigenvectors of path graph
1:
2:
3:
4:
17:
![Page 26: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/26.jpg)
Drawing of the graph using v3, v4
Plot vertex i at
-0.06 -0.04 -0.02 0 0.02 0.04 0.06-0.06
-0.04
-0.02
0
0.02
0.04
0.06
![Page 27: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/27.jpg)
Spectral graph coloring from high eigenvectors
Embedding of dodecahedron by 19th and 20th eigvecs.
Coloring 3-colorable random graphs [Alon-Kahale ’97]
![Page 28: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/28.jpg)
Spectral graph drawing: FEM justification
If apply finite element method to solve
Laplace’s equation in the plane
with a Delaunay triangulation
Would get graph Laplacian,
but with some weights on edges
Fundamental solutions are x and y coordinates
(see Strang’s Introduction to Applied Mathematics)
![Page 29: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/29.jpg)
Isomorphism testing
1. different eigenvalues -> non-isomorphic
2. If each vertex distinct in spectral embedding, just need to line up embeddings.
Each eigvec determined up to sign.
![Page 30: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/30.jpg)
Isomorphism testing
λ2 = λ3, eigvecs determined up to rotation
![Page 31: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/31.jpg)
Isomorphism testing
λ2 = λ3, eigvecs determined up to rotation
Distinguish bynorm in embedding
![Page 32: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/32.jpg)
Isomorphism testing: difficulties
2. If λi has a high dimensional space, eigvecsonly determined up to basis rotations
1. Many vertices can map to same place in spectral embedding, if only use few eigenvectors.
3. Some pairs have an exponential number of isomorphisms.
Ex.: Strongly regular graphs with only 3 eigenvalues,of multiplicities 1, (n-1)/2 and (n-1)/2
![Page 33: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/33.jpg)
Isomorphism testing: success[Babai-Grigoryev-Mount ‘82]
If each eigenvalue has multiplicity O(1), can test in polynomial time.
Ideas:Partition vertices into classes by norms in embeddings.Refine partitions using other partitions.Use vertex classes to split eigenspaces.
Use computational group theory to fuse information,and produce description of all isomorphisms.
![Page 34: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/34.jpg)
Random Walks
![Page 35: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/35.jpg)
Random walks and PageRank
Adjacency matrix of directed graph:
Walk transition matrix:
Walk distribution at time t:
PageRank vector p:
Eigenvector of Eigenvalue 1
![Page 36: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/36.jpg)
Random walks and PageRank
PageRank vector p:
Linear algebra issues:W is not symmetric, not similar to symmetric,does not necessarily have n eigenvalues
If no nodes of out-degree 0,Perron-Frobenius Theorem:
Guarantees a unique, positive eigevec p ofeigenvalue 1.
Is there a theoretically interesting spectral theory?
![Page 37: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/37.jpg)
Kleinberg and the singular vectors
Consider eigenvectors of largest eigenvalue of
and
Are left and right singular values of A.
Always exist.
Usually, a more useful theory than eigenvectors,when not symmetric.
(see Strang’s Intro. to Linear Algebra)
![Page 38: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/38.jpg)
Random walks on Undirected Graphs
Trivial PageRank Vector:
Not symmetric, but similiar to symmetrized walk matrix
W and S have same eigvals,
![Page 39: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/39.jpg)
Random walk converges at rate 1/1-λn-1
For lazy random walk (stay put with prob ½):
Where π is the stable distribution
For symmetric S
![Page 40: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/40.jpg)
Normalized Laplacian [Chung]
If consider 1-λn-1 should look at
Relationship to cuts:
![Page 41: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/41.jpg)
Cheeger’s Inequality (Jerrum-Sinclair ‘89)(Alon-Milman ‘85, Diaconis-Stroock ‘91)
![Page 42: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/42.jpg)
Cheeger’s Inequality (Jerrum-Sinclair ‘89)(Alon-Milman ‘85, Diaconis-Stroock ‘91)
Can find the cut by looking at
for some t
![Page 43: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/43.jpg)
Can find the cut by looking at
for some t
Only need approximate eigenvector(Mihail ’89)
Guarantee
Lanczos era.
![Page 44: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/44.jpg)
Normalized Cut
Alternative definition of conductance [Lovasz ’96 (?)]
This way, is a relaxation [see Hagen-Kahng ’92].
Equivalent to Normalized Cut [Shi-Malik ’00]
![Page 45: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/45.jpg)
Spectral Image Segmentation (Shi-Malik ‘00)
200 400 600 800 1000 1200
100
200
300
400
500
600
700
800
900
![Page 46: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/46.jpg)
Spectral Image Segmentation (Shi-Malik ‘00)
2 4 6 8 10 12 14 16
2
4
6
8
10
12
![Page 47: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/47.jpg)
Spectral Image Segmentation (Shi-Malik ‘00)
2 4 6 8 10 12 14 16
2
4
6
8
10
12
![Page 48: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/48.jpg)
Spectral Image Segmentation (Shi-Malik ‘00)
2 4 6 8 10 12 14 16
2
4
6
8
10
12
![Page 49: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/49.jpg)
Spectral Image Segmentation (Shi-Malik ‘00)
2 4 6 8 10 12 14 16
2
4
6
8
10
12
edge weight
![Page 50: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/50.jpg)
The second eigenvector
50 100 150 200 250 300
50
100
150
200
![Page 51: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/51.jpg)
Second Eigenvector’s sparsest cut
50 100 150 200 250 300
50
100
150
200
![Page 52: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/52.jpg)
Third Eigenvector
50 100 150 200 250 300
50
100
150
200
50 100 150 200 250 300
50
100
150
200
![Page 53: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/53.jpg)
Fourth Eigenvector
50 100 150 200 250 300
50
100
150
200
50 100 150 200 250 300
50
100
150
200
![Page 54: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/54.jpg)
Perspective on Spectral Image Segmentation
Ignoring a lot we know about images.
On non-image data, gives good intuition.
Can we fuse with what we know about images?
Generally, can we fuse with other knowledge?
What about better cut algorithms?
![Page 55: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/55.jpg)
Improvement by Miller and Tolliver ’06
50 100 150 200 250 300
50
100
150
200
50 100 150 200 250 300
50
100
150
200
![Page 56: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/56.jpg)
Improvement by Miller and Tolliver ’06
Idea: re-weight (i,j) by
Actually, re-weight by
Prove: as iterate λ2 −> 0, get 2 components
![Page 57: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/57.jpg)
One approach to fusing: Dirichlet Eigenvalues
Fixing boundary values to zero [Chung-Langlands ’96]
Fixing boundary values to non-zero. [Grady ’06] Dominant mode by solving linear equation:
computing electrical flow in resistor network
![Page 58: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/58.jpg)
Analysis of Spectral Partitioning
Finite Element Meshes (eigvals right) [S-Teng ’07]
Planted partitions (eigvecs right) [McSherry ‘01]
p q
q p
}}}}
A
BA
B
ProbA-A edge = pB-B edge = pA-B edge = q
q < p
![Page 59: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/59.jpg)
Other planted problems
Finding cn1/2 clique in random graph[Alon-Krivelevich-Sudakov ’98]
Color random sparse k-colorable graph[Alon-Kahale ’97]
Asymmetric block structure (LSI and HITS)[Azar-Fiat-Karlin-McSherry-Saia ’01]
Partitioning with widely varying degrees[Dasgupta-Hopcroft-McSherry ’04]
![Page 60: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/60.jpg)
Planted problem analysis
Small perturbations don’t change eigenvalues too much.
Eigenvectors stable too, if well-separated from others.
Understand eigenvalues of random matrices[Furedi-Komlos ’81, Alon-Krivelevich-Vu ’01, Vu ‘05]
p q
q p
Sampled A as perturbation of
![Page 61: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/61.jpg)
Distribution of eigenvalues of Random Graphs
-10 -5 0 5 10 15 20 250
2000
4000
6000
8000
10000
12000Histogram ofeigvals ofrandom40-by-40adjacencymatrices
![Page 62: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/62.jpg)
-10 -5 0 5 10 15 20 250
2000
4000
6000
8000
10000
12000
Distribution of eigenvalues of Random Graphs
Predicted curve: Wigner’s Semi-Circle Law
Histogram ofeigvals ofrandom40-by-40adjacencymatrices
![Page 63: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/63.jpg)
0 500 1000 1500 2000 2500-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
Eigenvalue distributions
Eigenvalues of walk matrix of 50-by-50 grid graph
Number greater than 1-ε proportional to ε
![Page 64: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/64.jpg)
0 500 1000 1500 2000 2500-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
Eigenvalue distributions
Number greater than 1-ε proportional to ε
All these go to zerowhen take powers
Eigenvalues of walk matrix of 50-by-50 grid graph
![Page 65: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/65.jpg)
Compression of powers of graphs[Coifman, Lafon, Lee, Maggioni, Nadler, Warner, Zucker ’05]
If most eigenvalues of A and W bounded from 1.Most eigenvalues of At very small.Can approximate At by low-rank matrix.
Build wavelets bases on graphs.Solve linear equations and compute eigenvectors.
Make rigorous by taking graph fromdiscretization of manifold
![Page 66: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/66.jpg)
Discretizing Manifold
edge weight
![Page 67: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/67.jpg)
Eigenvalue distributions
Eigenvalues of path graph on 10k nodes
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
Number greater than 1-ε proportional to
![Page 68: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/68.jpg)
Proof: can choose vertices to collapse so thatconductance becomes at least (like adding an expander on those nodes).
New graph has all eigvals at most 1-ε in abs value.Is rank change, so by Courant-Fischer
Theorem:If bounded degree, number eigenvalues greater than 1-ε is
Theorem: Eigenvalue distributions
![Page 69: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/69.jpg)
Eigenvalue distributions of planar graphs?
For planar graphs, Colan de Verdiere’s results imply
How big must the gap be?Must other gaps exist?
![Page 70: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/70.jpg)
Computation
Not rational, so only approximate
If λ exactly an eigenvalue,eigvecs = Null(A – λI)
If λ close to just one eigenvalue λi
If λi close to λi+1 is like λi = λi+1vi and vi+1 can rotate with each other
![Page 71: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/71.jpg)
General Symmetric Matrices
Locate any eigval in time
1.Orthogonal similarity transform to tri-diagonalin time by elimination algorithm.
2. Given tri-diagonal matrix, count number eigenvalues in any intervalin time
3. Do binary search to locate eigenvalue
Locate eigenvector: steps on tri-diagonal,time to map back to A
![Page 72: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/72.jpg)
Largest eigenvectors by power method
Apply A to random vector r:
In iters, expect x such that
Using Lanczos, expect iters(better polynomial)
![Page 73: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/73.jpg)
Smallest eigenvectors by inverse power method
Apply L-1 to random vector r orthogonal to
In iters, expect x such that
Compute in time [STeng04]if planar, in time [Koutis-Miller 06]
![Page 74: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/74.jpg)
Sparsification
Key to fast computation.
Replace A by sparse B for which
Generalized eigenvalues provide notion ofapproximation in graphs.
![Page 75: Spectral Graph Theory and its Applicationscs- Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline](https://reader031.vdocument.in/reader031/viewer/2022021512/5af5928e7f8b9a8d1c8dc4c7/html5/thumbnails/75.jpg)
Questions
Cheeger’s inequality for other physical problems?
How to incorporate other data into spectral methods?
Make multilevel coarsening rigorous.
What can we do with boundary conditions?
What about generalized eigenvalue problems?