communication complexity
play

Communication Complexity Lecture 23 Computing with remote inputs - PowerPoint PPT Presentation

Communication Complexity Lecture 23 Computing with remote inputs 1 Communication Complexity 2 Communication Complexity Setting 2 Communication Complexity Setting Alice wants to compute f(x,y) 2 Communication Complexity Setting


  1. Lower-Bounding χ (f) If a fooling set of size S, no two input-pairs from S can be on the same tile in a monochromatic tiling χ (f) ≥ |S| for every fooling set S Rank lower-bound χ (f) ≥ Rank(M f ) Discrepancy lower-bound χ (f) ≥ Discrepancy(f) 12

  2. Rank(M) 13

  3. Rank(M) Rank of a matrix 13

  4. Rank(M) Rank of a matrix Maximum number of linearly independent rows (or equivalently, columns) 13

  5. Rank(M) Rank of a matrix Maximum number of linearly independent rows (or equivalently, columns) Linear independence: operations in a field 13

  6. Rank(M) Rank of a matrix Maximum number of linearly independent rows (or equivalently, columns) Linear independence: operations in a field Rank-r matrix: after row & column reductions D (mxn) diagonal, with r 1’ s, rest 0’ s. M = UDV 13

  7. Rank(M) Rank of a matrix Maximum number of linearly independent rows (or equivalently, columns) Linear independence: operations in a field Rank-r matrix: after row & column reductions D (mxn) diagonal, with r 1’ s, rest 0’ s. M = UDV Rank(M) ≤ r, iff M can be written as sum of ≤ r rank 1 matrices 13

  8. Rank(M) Rank of a matrix Maximum number of linearly independent rows (or equivalently, columns) Linear independence: operations in a field Rank-r matrix: after row & column reductions D (mxn) diagonal, with r 1’ s, rest 0’ s. M = UDV Rank(M) ≤ r, iff M can be written as sum of ≤ r rank 1 matrices M = UDV = Σ i ≤ r D ii U i(mx1) V i(1xn) = Σ i ≤ r B i , where Rank(B i )=1 13

  9. Rank(M) Rank of a matrix Maximum number of linearly independent rows (or equivalently, columns) Linear independence: operations in a field Rank-r matrix: after row & column reductions D (mxn) diagonal, with r 1’ s, rest 0’ s. M = UDV Rank(M) ≤ r, iff M can be written as sum of ≤ r rank 1 matrices M = UDV = Σ i ≤ r D ii U i(mx1) V i(1xn) = Σ i ≤ r B i , where Rank(B i )=1 If M = Σ i ≤ r B i = UDV, Rank(M) ≤ min{Rank(U),Rank(D),Rank(V)} ≤ Rank(D) = r 13

  10. χ (f) ≥ Rank(M f ) 14

  11. χ (f) ≥ Rank(M f ) If M = Σ i ≤ r B i with Rank(B i )=1, then Rank(M) ≤ r 14

  12. χ (f) ≥ Rank(M f ) If M = Σ i ≤ r B i with Rank(B i )=1, then Rank(M) ≤ r M f = Σ i ≤ χ (f) Tile i , where Tile i has a monochromatic rectangle and 0’ s elsewhere 14

  13. χ (f) ≥ Rank(M f ) If M = Σ i ≤ r B i with Rank(B i )=1, then Rank(M) ≤ r M f = Σ i ≤ χ (f) Tile i , where Tile i has a monochromatic rectangle and 0’ s elsewhere Rank(Tile i )=1 14

  14. χ (f) ≥ Rank(M f ) If M = Σ i ≤ r B i with Rank(B i )=1, then Rank(M) ≤ r M f = Σ i ≤ χ (f) Tile i , where Tile i has a monochromatic rectangle and 0’ s elsewhere Rank(Tile i )=1 Rank(M f ) ≤ χ (f) 14

  15. χ (f) ≥ Rank(M f ) If M = Σ i ≤ r B i with Rank(B i )=1, then Rank(M) ≤ r M f = Σ i ≤ χ (f) Tile i , where Tile i has a monochromatic rectangle and 0’ s elsewhere Rank(Tile i )=1 Rank(M f ) ≤ χ (f) CC(f) ≥ log( χ (f)) ≥ log(Rank(M f )) 14

  16. Discrepancy 15

  17. Discrepancy Discrepancy of a 0-1 matrix 15

  18. Discrepancy Discrepancy of a 0-1 matrix max “imbalance” in any rectangle 15

  19. Discrepancy Discrepancy of a 0-1 matrix max “imbalance” in any rectangle Imbalance = | #1’ s - #0’ s | 15

  20. Discrepancy Discrepancy of a 0-1 matrix max “imbalance” in any rectangle Imbalance = | #1’ s - #0’ s | Disc(M) = 1/(mn) max rect imbalance(rect) 15

  21. Discrepancy Discrepancy of a 0-1 matrix max “imbalance” in any rectangle Imbalance = | #1’ s - #0’ s | Disc(M) = 1/(mn) max rect imbalance(rect) χ (f) ≥ 1/Disc(M f ) 15

  22. Discrepancy Discrepancy of a 0-1 matrix max “imbalance” in any rectangle Imbalance = | #1’ s - #0’ s | Disc(M) = 1/(mn) max rect imbalance(rect) χ (f) ≥ 1/Disc(M f ) Disc(M f ) ≥ 1/(mn) (size of largest monochromatic tile) 15

  23. Discrepancy Discrepancy of a 0-1 matrix max “imbalance” in any rectangle Imbalance = | #1’ s - #0’ s | Disc(M) = 1/(mn) max rect imbalance(rect) χ (f) ≥ 1/Disc(M f ) Disc(M f ) ≥ 1/(mn) (size of largest monochromatic tile) χ (f) ≥ (mn)/(size of largest monochromatic tile) 15

  24. CC Lower-bounds Summary 16

  25. CC Lower-bounds Summary CC(f) ≥ log(#transcripts) 16

  26. CC Lower-bounds Summary CC(f) ≥ log(#transcripts) Tiling Lower-bound: #transcripts ≥ χ (f) 16

  27. CC Lower-bounds Summary CC(f) ≥ log(#transcripts) Tiling Lower-bound: #transcripts ≥ χ (f) Both fairly tight: CC(f) = O( log 2 ( χ (f)) ) 16

  28. CC Lower-bounds Summary CC(f) ≥ log(#transcripts) Tiling Lower-bound: #transcripts ≥ χ (f) Both fairly tight: CC(f) = O( log 2 ( χ (f)) ) To lower-bound χ (f): fooling-set, rank, 1/Disc 16

  29. CC Lower-bounds Summary CC(f) ≥ log(#transcripts) Tiling Lower-bound: #transcripts ≥ χ (f) Both fairly tight: CC(f) = O( log 2 ( χ (f)) ) To lower-bound χ (f): fooling-set, rank, 1/Disc χ (f) ≥ |max fooling-set| ≥ (Rank(M f )) 2 16

  30. CC Lower-bounds Summary CC(f) ≥ log(#transcripts) Tiling Lower-bound: #transcripts ≥ χ (f) Both fairly tight: CC(f) = O( log 2 ( χ (f)) ) To lower-bound χ (f): fooling-set, rank, 1/Disc χ (f) ≥ |max fooling-set| ≥ (Rank(M f )) 2 1/Discrepancy lower-bounds can be very loose 16

  31. CC Lower-bounds Summary CC(f) ≥ log(#transcripts) Tiling Lower-bound: #transcripts ≥ χ (f) Both fairly tight: CC(f) = O( log 2 ( χ (f)) ) To lower-bound χ (f): fooling-set, rank, 1/Disc χ (f) ≥ |max fooling-set| ≥ (Rank(M f )) 2 1/Discrepancy lower-bounds can be very loose Conjecture: Rank(M f ) (and hence fooling set) is fairly tight 16

  32. CC Lower-bounds Summary CC(f) ≥ log(#transcripts) Tiling Lower-bound: #transcripts ≥ χ (f) Both fairly tight: CC(f) = O( log 2 ( χ (f)) ) To lower-bound χ (f): fooling-set, rank, 1/Disc χ (f) ≥ |max fooling-set| ≥ (Rank(M f )) 2 1/Discrepancy lower-bounds can be very loose Conjecture: Rank(M f ) (and hence fooling set) is fairly tight i.e., CC(f) = O(polylog(Rank(M f )) 16

  33. Many Variants 17

  34. Many Variants Randomized protocols: significant savings in expectation 17

  35. Many Variants Randomized protocols: significant savings in expectation Non-deterministic: Alice and Bob are non-deterministic. “Communication” now includes shared guess 17

Recommend


More recommend