communication complexity with small advantage
play

Communication Complexity with Small Advantage Thomas Watson - PowerPoint PPT Presentation

Communication Complexity with Small Advantage Thomas Watson University of Memphis Communication complexity F : { 0 , 1 } n { 0 , 1 } n { 0 , 1 } Communication complexity F : { 0 , 1 } n { 0 , 1 } n { 0 , 1 } (Alice: x ) (Bob: y )


  1. Communication Complexity with Small Advantage Thomas Watson University of Memphis

  2. Communication complexity F : { 0 , 1 } n × { 0 , 1 } n → { 0 , 1 }

  3. Communication complexity F : { 0 , 1 } n × { 0 , 1 } n → { 0 , 1 } (Alice: x ) (Bob: y )

  4. Communication complexity F : { 0 , 1 } n × { 0 , 1 } n → { 0 , 1 } (Alice: x ) − − − − − − − − − → (Bob: y )

  5. Communication complexity F : { 0 , 1 } n × { 0 , 1 } n → { 0 , 1 } (Alice: x ) − − − − − − − − − → (Bob: y ) ← − − − − − − − − −

  6. Communication complexity F : { 0 , 1 } n × { 0 , 1 } n → { 0 , 1 } (Alice: x ) − − − − − − − − − → (Bob: y ) ← − − − − − − − − − − − − − − − − − − →

  7. Communication complexity F : { 0 , 1 } n × { 0 , 1 } n → { 0 , 1 } (Alice: x ) − − − − − − − − − → (Bob: y ) ← − − − − − − − − − − − − − − − − − − → ← − − − − − − − − −

  8. Communication complexity F : { 0 , 1 } n × { 0 , 1 } n → { 0 , 1 } (Alice: x ) − − − − − − − − − → (Bob: y ) ← − − − − − − − − − − − − − − − − − − → ← − − − − − − − − − − − − − − − − − − →

  9. Communication complexity F : { 0 , 1 } n × { 0 , 1 } n → { 0 , 1 } (Alice: x ) − − − − − − − − − → (Bob: y ) ← − − − − − − − − − − − − − − − − − − → ← − − − − − − − − − − − − − − − − − − → F ( x , y ) F ( x , y )

  10. Communication complexity F : { 0 , 1 } n × { 0 , 1 } n → { 0 , 1 } (Alice: x ) − − − − − − − − − → (Bob: y ) ← − − − − − − − − − − − − − − − − − − → ← − − − − − − − − − − − − − − − − − − → F ( x , y ) F ( x , y ) Randomized protocols:

  11. Communication complexity F : { 0 , 1 } n × { 0 , 1 } n → { 0 , 1 } (Alice: x ) − − − − − − − − − → (Bob: y ) ← − − − − − − − − − − − − − − − − − − → ← − − − − − − − − − − − − − − − − − − → F ( x , y ) F ( x , y ) Randomized protocols: ◮ Correctness: ∀ ( x , y ) : P [output is F ( x , y )] ≥ 3 / 4

  12. Communication complexity F : { 0 , 1 } n × { 0 , 1 } n → { 0 , 1 } (Alice: x ) − − − − − − − − − → (Bob: y ) ← − − − − − − − − − − − − − − − − − − → ← − − − − − − − − − − − − − − − − − − → F ( x , y ) F ( x , y ) Randomized protocols: ◮ Correctness: ∀ ( x , y ) : P [output is F ( x , y )] ≥ 3 / 4 ◮ Cost: max ( x , y ) , random outcomes (# bits communicated)

  13. Communication complexity F : { 0 , 1 } n × { 0 , 1 } n → { 0 , 1 } (Alice: x ) − − − − − − − − − → (Bob: y ) ← − − − − − − − − − − − − − − − − − − → ← − − − − − − − − − − − − − − − − − − → F ( x , y ) F ( x , y ) Randomized protocols: ◮ Correctness: ∀ ( x , y ) : P [output is F ( x , y )] ≥ 3 / 4 ◮ Cost: max ( x , y ) , random outcomes (# bits communicated) ◮ Complexity: R( F ) = min correct protocols (cost)

  14. Classic results

  15. Classic results R( Inner-Product ) = Θ( n ) R( Set-Intersection ) = Θ( n ) R( Gap-Hamming ) = Θ( n )

  16. Classic results R( Inner-Product ) = Θ( n ) R( Set-Intersection ) = Θ( n ) R( Gap-Hamming ) = Θ( n ) R : success probability ≥ 3 / 4

  17. Classic results R( Inner-Product ) = Θ( n ) R( Set-Intersection ) = Θ( n ) R( Gap-Hamming ) = Θ( n ) R : success probability ≥ 3 / 4 Small advantage: R 1 / 2+ ǫ : success probability ≥ 1 / 2 + ǫ

  18. Classic results — revisited (unless ǫ ≤ 2 − Ω( n ) ) R 1 / 2+ ǫ ( Inner-Product ) = Θ( n ) R 1 / 2+ ǫ ( Set-Intersection ) = Θ( ǫ · n ) = Θ( ǫ 2 · n ) R 1 / 2+ ǫ ( Gap-Hamming )

  19. Classic results — revisited (unless ǫ ≤ 2 − Ω( n ) ) R 1 / 2+ ǫ ( Inner-Product ) = Θ( n ) R 1 / 2+ ǫ ( Set-Intersection ) = Θ( ǫ · n ) = Θ( ǫ 2 · n ) R 1 / 2+ ǫ ( Gap-Hamming ) . . . other functions?

  20. Climbing the polynomial hierarchy NP: R 1 / 2+ ǫ ( Set-Intersection ) = Θ( ǫ · n )

  21. Climbing the polynomial hierarchy NP: R 1 / 2+ ǫ ( Set-Intersection ) = Θ( ǫ · n ) [Braverman–Moitra STOC’13, G¨ o¨ os–Watson RANDOM’14] (information complexity) (corruption)

  22. Climbing the polynomial hierarchy NP: R 1 / 2+ ǫ ( Set-Intersection ) = Θ( ǫ · n ) [Braverman–Moitra STOC’13, G¨ o¨ os–Watson RANDOM’14] (information complexity) (corruption) Σ 2 P , Π 2 P:

  23. Climbing the polynomial hierarchy NP: R 1 / 2+ ǫ ( Set-Intersection ) = Θ( ǫ · n ) [Braverman–Moitra STOC’13, G¨ o¨ os–Watson RANDOM’14] (information complexity) (corruption) Σ 2 P , Π 2 P: R 1 / 2+ ǫ ( Tribes ) = Θ( ǫ · n )

  24. Climbing the polynomial hierarchy NP: R 1 / 2+ ǫ ( Set-Intersection ) = Θ( ǫ · n ) [Braverman–Moitra STOC’13, G¨ o¨ os–Watson RANDOM’14] (information complexity) (corruption) Σ 2 P , Π 2 P: R 1 / 2+ ǫ ( Tribes ) = Θ( ǫ · n ) Higher levels? (read-once AC 0 formulas)

  25. Climbing the polynomial hierarchy NP: R 1 / 2+ ǫ ( Set-Intersection ) = Θ( ǫ · n ) [Braverman–Moitra STOC’13, G¨ o¨ os–Watson RANDOM’14] (information complexity) (corruption) Σ 2 P , Π 2 P: R 1 / 2+ ǫ ( Tribes ) = Θ( ǫ · n ) Higher levels? (read-once AC 0 formulas) Constant advantage: well-understood [Jayram–Kopparty–Raghavendra/Leonardos–Saks CCC’09]

  26. Climbing the polynomial hierarchy NP: R 1 / 2+ ǫ ( Set-Intersection ) = Θ( ǫ · n ) [Braverman–Moitra STOC’13, G¨ o¨ os–Watson RANDOM’14] (information complexity) (corruption) Σ 2 P , Π 2 P: R 1 / 2+ ǫ ( Tribes ) = Θ( ǫ · n ) Higher levels? (read-once AC 0 formulas) Constant advantage: well-understood [Jayram–Kopparty–Raghavendra/Leonardos–Saks CCC’09] Small advantage: open

  27. Function definitions Set-Intersection : x y x y x y x y x y 1 1 2 2 3 3 4 4 5 5

  28. Function definitions Set-Intersection : x y x y x y x y x y 1 1 2 2 3 3 4 4 5 5 Tribes : n n x y x y x y x y x y x y x y 7 x y x y 1 1 2 2 3 3 4 4 5 5 6 6 7 8 8 9 9

  29. What’s known about Tribes ? R( Tribes ) = Θ( n )

  30. What’s known about Tribes ? R( Tribes ) = Θ( n ) [Jayram–Kumar–Sivakumar STOC’03, Harsha–Jain FSTTCS’13] (information complexity) (smooth rectangle bound)

  31. What’s known about Tribes ? R( Tribes ) = Θ( n ) [Jayram–Kumar–Sivakumar STOC’03, Harsha–Jain FSTTCS’13] (information complexity) (smooth rectangle bound) R 1 / 2+ ǫ ( Tribes ) = ??

  32. What’s known about Tribes ? R( Tribes ) = Θ( n ) [Jayram–Kumar–Sivakumar STOC’03, Harsha–Jain FSTTCS’13] (information complexity) (smooth rectangle bound) R 1 / 2+ ǫ ( Tribes ) = ?? [G¨ o¨ os–Watson] trick: R 1 / 2+ ǫ ≥ Ω( ǫ · corruption bound)

  33. What’s known about Tribes ? R( Tribes ) = Θ( n ) [Jayram–Kumar–Sivakumar STOC’03, Harsha–Jain FSTTCS’13] (information complexity) (smooth rectangle bound) R 1 / 2+ ǫ ( Tribes ) = ?? [G¨ o¨ os–Watson] trick: R 1 / 2+ ǫ ≥ Ω( ǫ · corruption bound) ◮ Doesn’t work for Tribes : corruption bound ≈ √ n

  34. What’s known about Tribes ? R( Tribes ) = Θ( n ) [Jayram–Kumar–Sivakumar STOC’03, Harsha–Jain FSTTCS’13] (information complexity) (smooth rectangle bound) R 1 / 2+ ǫ ( Tribes ) = ?? [G¨ o¨ os–Watson] trick: R 1 / 2+ ǫ ≥ Ω( ǫ · corruption bound) ◮ Doesn’t work for Tribes : corruption bound ≈ √ n ?? Similar trick: R 1 / 2+ ǫ ≥ Ω( ǫ · smooth rectangle bound) ??

  35. What’s known about Tribes ? R( Tribes ) = Θ( n ) [Jayram–Kumar–Sivakumar STOC’03, Harsha–Jain FSTTCS’13] (information complexity) (smooth rectangle bound) R 1 / 2+ ǫ ( Tribes ) = ?? [G¨ o¨ os–Watson] trick: R 1 / 2+ ǫ ≥ Ω( ǫ · corruption bound) ◮ Doesn’t work for Tribes : corruption bound ≈ √ n ?? Similar trick: R 1 / 2+ ǫ ≥ Ω( ǫ · smooth rectangle bound) ?? ◮ Not true in general

  36. Our approach for Tribes Information complexity:

  37. Our approach for Tribes Information complexity: ◮ Ω(1)-advantage for Tribes [JKS’03]

  38. Our approach for Tribes Information complexity: ◮ Ω(1)-advantage for Tribes [JKS’03] ◮ ǫ -advantage for Set-Inter [BM’13]

  39. Our approach for Tribes Information complexity: ◮ Ω(1)-advantage for Tribes [JKS’03] ◮ ǫ -advantage for Set-Inter [BM’13] ◮ Combine?

  40. Our approach for Tribes Information complexity: ◮ Ω(1)-advantage for Tribes [JKS’03] ◮ ǫ -advantage for Set-Inter [BM’13] ◮ Combine? 4-step approach:

Recommend


More recommend