improved bounds for mcmc sampling colorings of g n d n
play

Improved bounds for MCMC sampling colorings of G ( n , d / n ) - PowerPoint PPT Presentation

Improved bounds for MCMC sampling colorings of G ( n , d / n ) Charis Efthymiou efthymiou@gmail.com Goethe University, Frankfurt Joint work with: T. Hayes, D. Stefankovi c and E. Vigoda Workshop on Local Algorithms MIT Boston, June,


  1. Block Construction Weights [Efthymiou (2014)] • Each vertex u of degree deg ( u ) is assigned weight � (1 + γ ) − 1 deg ( u ) ≤ (1 + ǫ ) d W ( u ) = d c · deg ( u ) otherwise • Every path L is assigned weight � u ∈ L W ( u ) “Break Points” ln n Γ( v ) := set of paths of length at most d 2 / 5 that emanate from v . For a break-point v , we have

  2. Block Construction Weights [Efthymiou (2014)] • Each vertex u of degree deg ( u ) is assigned weight � (1 + γ ) − 1 deg ( u ) ≤ (1 + ǫ ) d W ( u ) = d c · deg ( u ) otherwise • Every path L is assigned weight � u ∈ L W ( u ) “Break Points” ln n Γ( v ) := set of paths of length at most d 2 / 5 that emanate from v . For a break-point v , we have �� � max W ( u ) ≤ 1 . L ∈ Γ( v ) u ∈ L

  3. How do the Blocks look like

  4. How do the Blocks look like

  5. How do the Blocks look like Boundary of the block Consists only of break points.

  6. How do the Blocks look like Low degree “buffer” . . . between boundary vertices and a high degree vertex

  7. How do the Blocks look like . . . for the analysis the effect of high degrees disappears

  8. Proving Rapid Mixing

  9. Proving Rapid Mixing Path Coupling, [Bubley, Dyer 1997]

  10. Proving Rapid Mixing Path Coupling, [Bubley, Dyer 1997] Consider ( X t ), ( Y t ) such that X 0 ⊕ Y 0 = { w ∗ }

  11. Proving Rapid Mixing Path Coupling, [Bubley, Dyer 1997] Consider ( X t ), ( Y t ) such that X 0 ⊕ Y 0 = { w ∗ } For rapid mixing it suffices to have a coupling such that E [ dist ( X 1 , Y 1 ) | X 0 , Y 0 ] ≤ (1 − γ ) dist ( X 0 , Y 0 ) ,

  12. Proving Rapid Mixing Path Coupling, [Bubley, Dyer 1997] Consider ( X t ), ( Y t ) such that X 0 ⊕ Y 0 = { w ∗ } For rapid mixing it suffices to have a coupling such that E [ dist ( X 1 , Y 1 ) | X 0 , Y 0 ] ≤ (1 − γ ) dist ( X 0 , Y 0 ) , where � dist ( σ, τ ) = β ( u ) u ∈ σ ⊕ τ

  13. Distance between σ and τ

  14. Distance between σ and τ dist ( σ, τ ) depends on the block partition B .

  15. Distance between σ and τ dist ( σ, τ ) depends on the block partition B .

  16. Distance between σ and τ dist ( σ, τ ) depends on the block partition B .

  17. Distance between σ and τ A distance that counts the disagreeing edges between the blocks

  18. Distance between σ and τ A new distance metric Given G ( n , d / n ) and set of blocks B , for any two σ, τ � dist ( σ, τ ) = 1 { v ∈ σ ⊕ τ } deg out ( v ) v ∈ ∂ B

  19. Distance between σ and τ A new distance metric Given G ( n , d / n ) and set of blocks B , for any two σ, τ dist ( σ, τ ) = n 2 � � 1 { v ∈ σ ⊕ τ } deg out ( v ) + 1 { v ∈ σ ⊕ τ } v ∈ ∂ B v ∈ V \ ∂ B

  20. Distance between σ and τ A new distance metric Given G ( n , d / n ) and set of blocks B , for any two σ, τ dist ( σ, τ ) = n 2 � � 1 { v ∈ σ ⊕ τ } deg out ( v )+ 1 { v ∈ σ ⊕ τ } v ∈ ∂ B v ∈ V \ ∂ B

  21. The coupling

  22. The coupling B 2 B 0 B 3 B 1 B 4

  23. The coupling B 2 B 0 B 3 B 1 B 4

  24. The coupling B 2 B 0 B 3 B 1 B 4

  25. The coupling B 2 B 0 B 3 B 1 B 4

  26. The coupling B 2 B 0 B 3 B 1 B 4

  27. The coupling of X ( B ) and Y ( B )

  28. The coupling of X ( B ) and Y ( B ) • one vertex at a time

  29. The coupling of X ( B ) and Y ( B ) • one vertex at a time • pick a vertex next to a disagreement

  30. The coupling of X ( B ) and Y ( B ) • one vertex at a time • pick a vertex next to a disagreement

  31. The coupling of X ( B ) and Y ( B ) • one vertex at a time • pick a vertex next to a disagreement

  32. The coupling of X ( B ) and Y ( B ) • one vertex at a time • pick a vertex next to a disagreement

  33. The coupling of X ( B ) and Y ( B ) • one vertex at a time • pick a vertex next to a disagreement

  34. The coupling of X ( B ) and Y ( B ) • one vertex at a time • pick a vertex next to a disagreement

  35. The coupling of X ( B ) and Y ( B ) • one vertex at a time • pick a vertex next to a disagreement • disagreement probability 1 � deg ( v ) < k k − deg ( v ) ̺ v = 1 otherwise

  36. The coupling of X ( B ) and Y ( B ) • one vertex at a time • pick a vertex next to a disagreement • disagreement probability 1 � deg ( v ) < k k − deg ( v ) ̺ v = 1 otherwise • probability of the most likely color

  37. The coupling of X ( B ) and Y ( B ) • one vertex at a time • pick a vertex next to a disagreement • disagreement probability 1 � deg ( v ) < k k − deg ( v ) ̺ v = 1 otherwise • probability of the most likely color

  38. The coupling of X ( B ) and Y ( B ) • one vertex at a time • pick a vertex next to a disagreement • disagreement probability 1 � deg ( v ) < k k − deg ( v ) ̺ v = 1 otherwise • probability of the most likely color

  39. The coupling of X ( B ) and Y ( B ) • one vertex at a time • pick a vertex next to a disagreement • disagreement probability 1 � deg ( v ) < k k − deg ( v ) ̺ v = 1 otherwise • probability of the most likely color

  40. Rapid Mixing for k > 2 d

Recommend


More recommend