earthmover resilience testing in ordered structures
play

Earthmover resilience & testing in ordered structures Eldar - PowerPoint PPT Presentation

Earthmover resilience & testing in ordered structures Eldar Fischer Omri Ben-Eliezer Technion Tel-Aviv University Computational Complexity Conference 2018 UCSD, San Diago Property testing (RS96, GGR98) Meta problem: Given property P,


  1. Earthmover resilience & testing in ordered structures Eldar Fischer Omri Ben-Eliezer Technion Tel-Aviv University Computational Complexity Conference 2018 UCSD, San Diago

  2. Property testing (RS96, GGR98) Meta problem: Given property P, efficiently distinguish between • Objects that satisfy P • Objects that are far from satisfying P

  3. Graph property testing Definition: An ! -test for a property P is given query access to an unknown graph G on n vertices, and acts as follows. " satis'ies ) G is ! -far from P ACCEPT (with prob. 2/3) DON’T CARE REJECT (with prob. 2/3)

  4. Graph property testing Query = “Is there an edge between u and v?” (dense graph model) " satis'ies ) G is ! -far from P ACCEPT (with prob. 2/3) DON’T CARE REJECT (with prob. 2/3)

  5. Graph property testing ! -far = need to add/remove !" # edges in G to satisfy P. (dense graph model) $ satis)ies + G is ! -far from P ACCEPT (with prob. 2/3) DON’T CARE REJECT (with prob. 2/3)

  6. Graph property testing Definition: A property P is testable if it has an ! -test making " ! queries for any ! > 0 . Question (GGR98): Which graph properties are testable?

  7. Canonical tests • An ! -test is canonical if it queries a random induced subgraph and accepts/rejects only based on queried subgraph.

  8. Canonical tests • Theorem [AFKS00, GT03]: P testable ó P canonically testable Intuition: Original test makes ! queries Canonical test picks random 2! vertices, then “simulates” original test

  9. Tolerant testing [PRR’06] • Test is !, # -tolerant ( 0 ≤ & < ( ) if it acts as follows. • Motivation: Noisy input ) satis.ies 0 G is # -far from P G is δ −close to P ACCEPT (with prob. 2/3) REJECT (with prob. 2/3) DON’T CARE

  10. Tolerant testing [PRR’06] P is tolerantly testable ∀" ∃$ : P has a $, & -test making '(&) queries. + satis0ies 2 G is * -far from P G is δ −close to P ACCEPT (with prob. 2/3) REJECT (with prob. 2/3) DON’T CARE

  11. Distance estimation P is estimable ∀" ∀# : P has a " − #, " -test making &(#, ") queries. G is ) -far from P G is () − *) −close to P ACCEPT (with prob. 2/3) DON’T CARE REJECT (with prob. 2/3)

  12. Testing vs to Te tolerant te testing vs di distanc nce estima mation • Theorem [Fischer, Newman ’05]: For graph properties, P canonically testable P estimable G is ! -far from P " G is ! -far from P G is (! − δ) −close to satis'ies ) P REJECT ACCEPT DON’T CARE REJECT ACCEPT DON’T CARE

  13. Su Summary - gr graph proper erti ties es testability canonical testability (GT03) G is ! -far from P G is ! -far from P " " satis'ies ) satis'ies ) trivial ACCEPT ACCEPT DON’T CARE REJECT DON’T CARE REJECT estimability tolerant testability (FN05) G is ! -far from P G is ! -far from P G δ −close to P G (! − δ) −close to P (FN05) ACCEPT DON’T CARE REJECT ACCEPT DON’T CARE REJECT

  14. What about ordered structures? • Strings (1D) • Images (2D) AKA ordered matrices • Vertex-ordered graphs (2D) and hypergraphs • Hypercube (high-D): a different story...

  15. Image property testing Unknown !×! image # over fixed set of pixels Σ Query = “What is the color of pixel in location (i,j)?” # satis*ies , # is % -far from P ACCEPT (with prob. 2/3) DON’T CARE REJECT (with prob. 2/3)

  16. Image property testing Unknown !×! image # over fixed set of pixels Σ % -far = need to modify %& ' pixels in # to satisfy P # satis,ies . # is % -far from P ACCEPT (with prob. 2/3) DON’T CARE REJECT (with prob. 2/3)

  17. Image property testing canonical test = pick randomly ! rows and ! columns , query all pixels in intersection. queried pixel

  18. String property testing Query access to unknown string of length ! over fixed alphabet Σ . canonical test = pick randomly # elements and query them. queried element

  19. What about ordered structures? Do similar characterizations hold for ordered structures? • No , testability/estimability ⇏ canonical testability • Example: ”not containing three consecutive 1-s” in 0/1 strings. • No , testability ⇏ tolerant testability. [Fischer, Fortnow ‘05] • Properties based on codes & PCPPs. • Yes , for “global enough” properties. [This work]

  20. Earthmover resilience (strings) Flip locations of neighboring entries Flip operation : Definition: Earthmover distance between strings S and S’ is ! " #, # % = ' ( ) ⋅ min{ number of flips to create S’ from S , ∞ } Definition: Property P is earthmover resilient if ∃-: 0,1 → (0,1) s.t. String # String #′ satisfies String #′ is @ A B, B % ≤ -(D) satisfies P D -close to P

  21. Earthmover resilience (images) Flip locations of neighboring Flip operation : rows/columns Definition: Earthmover distance between images ! and !′ is ( # $ %, % = ) * ⋅ min{ number of flips to create ! ’ from ! , ∞ } Definition: Property P is earthmover resilient if ∃.: 0,1 → (0,1) s.t. image % image %′ satisfies Image %′ is > ? !, ! @ ≤ .(B) satisfies P B -close to P

  22. Which properties are earthmover resilient? • All unordered graph properties [trivial] • All hereditary properties of strings, images & ordered graphs [AKNS00, A B F17] 0 0 0 0 0 • Global visual properties of images 0 0 1 1 0 0 1 0 1 1 1 1 Monotonicity: • Convexity of the 1’s a hereditary property 1 1 1 1 0 • 1’s form a half plane 1 1 1 1 1 • [This work]: In general, all properties with sparse boundary between 1’s and 0’s. 0 0 0 0 0 1 1 1 1 0 1 1 Convex shape of 1’s 1 1 0 1 1 1 1 0 0 1 1 0 1 0

  23. Ea Earthmover resilience vs ca canonica cal testing [This work]: For string properties P , P earthmover P canonically resilient testable For image and ordered graph properties P , P earthmover P tolerantly P canonically resilient testable testable

  24. onical testing to es Ca Canon estim timatio tion [This work]: For image and ordered graph properties P , P canonically P (canonically) testable estimable Corollary [A B F17 + This work]: P (canonically) P hereditary estimable

  25. ER ER pr prope perties s are si similar to gr graph ph pr prope perties For earthmover resilient properties of images / ordered graphs: Tolerant estimability testability Canonical testability

  26. Warmup proof: ER ER canonical testing in binary y strings ER => piecewise canonical testing • Consider Interval partition of string into sufficiently many parts. • In each interval, make sufficiently many random queries to estimate number of 0’s and 1’s. • Due to ER, this gives good estimate for distance to P : 0 1 ∈3 VD(S, S 8 ) !"#$%&'( ), + ≈ min Where VD(S,S’) denotes average variation distance between the distributions of 0’s and 1’s in each interval.

  27. Warmup proof: ER ER canonical testing in binary y strings piecewise canonical testing => canonical testing • Interval partition can be approximated by • Picking sufficiently many random queries. • Partitioning them artificially into intervals. • Consequently, piecewise canonical tests can be simulated by canonical ones.

  28. Bits from the proof: Sz Szemerédi regularity y lemma [Szemerédi ‘75]: Any graph has an equipartition of size !(#) , so that almost all pairs of parts are # -regular. density D Pair is ' -regular if Size Size ( − * ≤ # density d ≥ #& ≥ #& for any pair of subsets of size ≥ #& Size N Size N

  29. Bits from the proof: ca canonica cal testing --> es estimation • High level idea - unordered case [Fischer Newman ‘05] • Step 1: If P is canonically testable, densities of small induced subgraphs among graphs satisfying P different from those of graphs far from P. • Step 2: regular partitions of graphs satisfying P differ from graphs far from P.

  30. Bits from the proof: ca canonica cal testing --> es estimation • High level idea - unordered case [Fischer Newman ‘05] • Step 1: If P is canonically testable, densities of small induced subgraphs among graphs satisfying P different from those of graphs far from P. • Step 2: regular partitions of graphs satisfying P differ from graphs far from P. • Step 3: Estimating which regular partitions a graph has - doable with constant number of queries. • Step 4: distance of G from P ≈ min distance of a regular partition for G from a regular partition for P .

  31. Bits from the proof: ca canonica cal testing --> es estimation • Our observation • Above scheme essentially works for multipartite graphs. • Given ordered graph ! , take interval partition of the vertices, effectively approximating ! by a multipartite graph.

  32. Bonus: Regular reducibility [Alon, Fischer, Newman, Shapira ‘06]: A graph property P is canonically testable P can be “described” using regular partitions [This work]: Same holds for images and ordered graphs.

  33. Towards a limit object? • Proofs in [ABF’17] and this work rely on interval partitioning . • A limit object ( graphon -like [BCLSSV05; LS08; BCLSV08]) for images and ordered graphs via interval partitioning?

Recommend


More recommend