the cover time of random walks on random geometric graphs
play

The cover time of random walks on random geometric graphs Colin - PowerPoint PPT Presentation

The cover time of random walks on random geometric graphs Colin Cooper Alan Frieze G = ( V , E ) is a connected graph. ( | V | = n , | E | = m ). For u V let C u be the expected time taken for a simple random walk W u on G starting at u , to


  1. The cover time of random walks on random geometric graphs Colin Cooper Alan Frieze

  2. G = ( V , E ) is a connected graph. ( | V | = n , | E | = m ). For u ∈ V let C u be the expected time taken for a simple random walk W u on G starting at u , to visit every vertex of G .

  3. G = ( V , E ) is a connected graph. ( | V | = n , | E | = m ). For u ∈ V let C u be the expected time taken for a simple random walk W u on G starting at u , to visit every vertex of G . The cover time C G of G is defined as C G = max u ∈ V C u .

  4. G = ( V , E ) is a connected graph. ( | V | = n , | E | = m ). For u ∈ V let C u be the expected time taken for a simple random walk W u on G starting at u , to visit every vertex of G . The cover time C G of G is defined as C G = max u ∈ V C u . C G ≤ 4 m ( n − 1 ) : Alelliunas,Karp,Lipton,Lovász,Rackoff (1979)

  5. G = ( V , E ) is a connected graph. ( | V | = n , | E | = m ). For u ∈ V let C u be the expected time taken for a simple random walk W u on G starting at u , to visit every vertex of G . The cover time C G of G is defined as C G = max u ∈ V C u . C G ≤ 4 m ( n − 1 ) : Alelliunas,Karp,Lipton,Lovász,Rackoff (1979) ( 1 − o ( 1 )) n ln n ≤ C G ≤ ( 1 + o ( 1 )) 4 27 n 3 : Feige (1995)

  6. Cooper and Frieze If p = c log n / n and c > 1 then w.h.p. � � c C G n , p ∼ c log n log n . c − 1

  7. Cooper and Frieze If p = c log n / n and c > 1 then w.h.p. � � c C G n , p ∼ c log n log n . c − 1 Let c > 1 and let x denote the solution in ( 0 , 1 ) of x = 1 − e − cx . Let X g be the giant component of cx ( 2 − x ) 4 ( cx − log c ) n ( log n ) 2 . G n , p , p = c / n . Then w.h.p. C X g ∼

  8. Cooper and Frieze If p = c log n / n and c > 1 then w.h.p. � � c C G n , p ∼ c log n log n . c − 1 Let c > 1 and let x denote the solution in ( 0 , 1 ) of x = 1 − e − cx . Let X g be the giant component of cx ( 2 − x ) 4 ( cx − log c ) n ( log n ) 2 . G n , p , p = c / n . Then w.h.p. C X g ∼ Let G n , r denote a random r -regular graph on vertex set [ n ] with r ≥ 3 then w.h.p. C G n , r ∼ r − 1 r − 2 n log n .

  9. Cooper and Frieze If p = c log n / n and c > 1 then w.h.p. � � c C G n , p ∼ c log n log n . c − 1 Let c > 1 and let x denote the solution in ( 0 , 1 ) of x = 1 − e − cx . Let X g be the giant component of cx ( 2 − x ) 4 ( cx − log c ) n ( log n ) 2 . G n , p , p = c / n . Then w.h.p. C X g ∼ Let G n , r denote a random r -regular graph on vertex set [ n ] with r ≥ 3 then w.h.p. C G n , r ∼ r − 1 r − 2 n log n . Let G m ( n ) denote a preferential attachment graph of average degree 2 m then w.h.p. 2 m C G m ∼ m − 1 n log n .

  10. Cooper and Frieze If p = c log n / n and c > 1 then w.h.p. � � c C G n , p ∼ c log n log n . c − 1 Let c > 1 and let x denote the solution in ( 0 , 1 ) of x = 1 − e − cx . Let X g be the giant component of cx ( 2 − x ) 4 ( cx − log c ) n ( log n ) 2 . G n , p , p = c / n . Then w.h.p. C X g ∼ Let G n , r denote a random r -regular graph on vertex set [ n ] with r ≥ 3 then w.h.p. C G n , r ∼ r − 1 r − 2 n log n . Let G m ( n ) denote a preferential attachment graph of average degree 2 m then w.h.p. 2 m C G m ∼ m − 1 n log n . Let D n , p denote a random digraph with independent edge probability p ). If p = c log n / n and c > 1 then w.h.p. � � c C D n , p ∼ c log n log n . c − 1

  11. Random geometric graph G = G ( d , r , n ) in d dimensions: Sample n points V independently and uniformly at random from [ 0 , 1 ] d . For each point x draw a ball D ( x , r ) of radius r about x . V ( G ) = V and E ( G ) = {{ v , w } : w � = v , w ∈ D ( v , r ) }

  12. Random geometric graph G = G ( d , r , n ) in d dimensions: Sample n points V independently and uniformly at random from [ 0 , 1 ] d . For each point x draw a ball D ( x , r ) of radius r about x . V ( G ) = V and E ( G ) = {{ v , w } : w � = v , w ∈ D ( v , r ) } For simplicity we replace [ 0 , 1 ] d by a torus.

  13. Avin and Ercal d = 2 Theorem C G = Θ( n log n ) w . h . p ..

  14. Avin and Ercal d = 2 Theorem C G = Θ( n log n ) w . h . p .. Cooper and Frieze d ≥ 3: Theorem � 1 / d � c log n Let c > 1 be constant, and let r = . Then w.h.p. Υ d n � � c C G ∼ T c = c log n log n . c − 1 Υ d is the volume of the unit ball in d dimensions.

  15. First Visit Time Lemma. Let π x = deg ( x ) denote the steady state for a random walk on G . 2 m

  16. First Visit Time Lemma. Let π x = deg ( x ) denote the steady state for a random walk on G . 2 m Let the mixing time T be defined so that u , x ∈ V | P ( t ) u ( x ) − π x | ≤ n − 3 . max

  17. First Visit Time Lemma. Let π x = deg ( x ) denote the steady state for a random walk on G . 2 m Let the mixing time T be defined so that u , x ∈ V | P ( t ) u ( x ) − π x | ≤ n − 3 . max Let R v be the expected number of visits by W v within time T .

  18. First Visit Time Lemma. Let π x = deg ( x ) denote the steady state for a random walk on G . 2 m Let the mixing time T be defined so that u , x ∈ V | P ( t ) u ( x ) − π x | ≤ n − 3 . max Let R v be the expected number of visits by W v within time T . Fix u , v ∈ V . For s ≥ T let A s ( v ) = {W u does not visit v in [ T , s ] }

  19. First Visit Time Lemma. Suppose that the connected graph G = ( V , E ) has n vertices and m edges.

  20. First Visit Time Lemma. Suppose that the connected graph G = ( V , E ) has n vertices and m edges. Let π x = deg ( x ) denote the steady state for a random walk on G . 2 m

  21. First Visit Time Lemma. Suppose that the connected graph G = ( V , E ) has n vertices and m edges. Let π x = deg ( x ) denote the steady state for a random walk on G . 2 m Let the mixing time T be defined so that u , x ∈ V | P ( t ) u ( x ) − π x | ≤ n − 3 . max

  22. First Visit Time Lemma. Suppose that the connected graph G = ( V , E ) has n vertices and m edges. Let π x = deg ( x ) denote the steady state for a random walk on G . 2 m Let the mixing time T be defined so that u , x ∈ V | P ( t ) u ( x ) − π x | ≤ n − 3 . max Fix u , v ∈ V . For s ≥ T let A s ( v ) = {W u does not visit v in [ T , s ] } We try to get a good estimate of Pr ( A s ( v )) .

  23. Write H ( s ) = F ( s ) R ( s ) where h t = Pr ( W u ( t ) = v ) and H ( s ) = � ∞ t = T h t s t r t = Pr ( W v ( t ) = v ) and R ( s ) = � ∞ t = 0 r t s t f t is the probability that the first visit of W u to v in the period [ T , T + 1 , . . . , ] occurs at step t , and F ( s ) = � ∞ t = T f t s t .

  24. Write H ( s ) = F ( s ) R ( s ) where h t = Pr ( W u ( t ) = v ) and H ( s ) = � ∞ t = T h t s t r t = Pr ( W v ( t ) = v ) and R ( s ) = � ∞ t = 0 r t s t f t is the probability that the first visit of W u to v in the period [ T , T + 1 , . . . , ] occurs at step t , and F ( s ) = � ∞ t = T f t s t . Note that � Pr ( A s ( v )) = f t . t > s

  25. For | z | ≤ 1 + λ, λ = 1 / KT z T 1 − z + o ( n − 2 ) R ( z ) = R T ( z ) + π v z T 1 − z + o ( n − 2 ) H ( z ) = π v

  26. Now write F ( z ) = H ( z ) R ( z ) = B ( z ) A ( z ) where for | z | ≤ 1 + λ π v z T + ( 1 − z ) R T ( z ) + o ( n − 2 ) A ( z ) = π v z T + o ( n − 2 ) B ( z ) =

  27. Now write F ( z ) = H ( z ) R ( z ) = B ( z ) A ( z ) where for | z | ≤ 1 + λ π v z T + ( 1 − z ) R T ( z ) + o ( n − 2 ) A ( z ) = π v z T + o ( n − 2 ) B ( z ) = A ( z ) has a zero at π v z 0 = 1 + R T ( 1 ) + o ( 1 ) .

  28. Now write F ( z ) = H ( z ) R ( z ) = B ( z ) A ( z ) A ( z ) has a zero at π v z 0 = 1 + R T ( 1 ) + o ( 1 ) . One can then show that F ( z ) = B ( z 0 ) / A ′ ( z 0 ) + g ( z ) z − z 0 where g ( z ) is analytic inside | z | ≤ 1 + λ .

  29. One can then show that F ( z ) = B ( z 0 ) / A ′ ( z 0 ) + g ( z ) z − z 0 where g ( z ) is analytic inside | z | ≤ 1 + λ . and so − B ( z 0 ) / A ′ ( z 0 ) + O (( 1 + λ ) − t ) f t = z t + 1 0 π v / R T ( 1 + π v / R T ) t + 1 + O (( 1 + λ ) − t ) . ∼

  30. Pr ( A s ( v )) = e − ( 1 + o ( 1 )) π v s / R v .

  31. Pr ( A s ( v )) = e − ( 1 + o ( 1 )) π v s / R v . Most difficult task now is to show that R v = 1 + o ( 1 ) for all v .

  32. Important Sub-Structure Whp there is an embedded grid Γ made up of heavy sub-cubes, each contain Θ( log n ) vertices. Edges of grid are sequences of heavy cubes, of same length. Vertices of G are within O ( 1 ) distance of Γ .

  33. Given the grid Γ it is easy to use a Canonical Paths argument to show that T = ˜ O ( n 2 / d )

  34. Given the grid Γ it is easy to use a Canonical Paths argument to show that T = ˜ O ( n 2 / d ) This estimate is not very good for d = 2. In this case it can be improved to T = O ( n / log n ) .

  35. Canonical Paths We need two basic results on mixing times. First let λ max be the second largest eigenvalue of the transition matrix P . Then, � 1 / 2 � π x | P ( t ) λ t u ( x ) − π x | ≤ max . π u Next, for each x � = y ∈ V let γ xy be a canonical path from x to y in G . Then, we have that λ max ≤ 1 − 1 ρ, where 1 � ρ = max π ( a ) π ( b ) | γ ab | , π ( x ) P ( x , y ) e = { x , y }∈ E ( G ) γ ab ∋ e and | γ ab | is the length of the canonical path γ ab from a to b .

  36. Consider the grid Γ .It will help to fix a collection of points x i in each red cube for i = 1 , 2 , . . . , M .

Recommend


More recommend