what do shannon type inequalities submodular width and
play

What do Shannon-type Inequalities, Submodular Width, and - PowerPoint PPT Presentation

Feb 02, 2024 •263 likes •1.59k views

What do Shannon-type Inequalities, Submodular Width, and Disjunctive Datalog have to do with one another? Mahmoud Abo Khamis 1 Hung Q. Ngo 1 Dan Suciu 1 , 2 1 LogicBlox Inc. 2 University of Washington PODS 2017 1/22 Contributions A Join

  1. Size Bounds: Input ◮ A full conjunctive query � Q ( A [ n ] ) :- R F ( A F ) F ∈E ◮ A [ n ] = { A 1 , . . . , A n } is the full set of attributes ◮ A F = { A f | f ∈ F ⊆ [ n ] } is a subset ◮ Degree Constraints ( DC ): deg F ( A Y | A X ) ≤ N Y | X , X ⊂ Y ⊆ F ∈ E ◮ Cardinality Constraints ( CC ): | R F | ≤ N F |∅ ◮ Functional Dependencies ( FD ): A X → A Y Bound Idea 6/22

  2. Size Bounds: Input ◮ A full conjunctive query � Q ( A [ n ] ) :- R F ( A F ) F ∈E ◮ A [ n ] = { A 1 , . . . , A n } is the full set of attributes ◮ A F = { A f | f ∈ F ⊆ [ n ] } is a subset ◮ Degree Constraints ( DC ): deg F ( A Y | A X ) ≤ N Y | X , X ⊂ Y ⊆ F ∈ E ◮ Cardinality Constraints ( CC ): | R F | ≤ N F |∅ ◮ Functional Dependencies ( FD ): A X → A Y Bound Idea log | Q | ≤ maximum h ( A 1 , . . . , A n ) 6/22

  3. Size Bounds: Input ◮ A full conjunctive query � Q ( A [ n ] ) :- R F ( A F ) F ∈E ◮ A [ n ] = { A 1 , . . . , A n } is the full set of attributes ◮ A F = { A f | f ∈ F ⊆ [ n ] } is a subset ◮ Degree Constraints ( DC ): deg F ( A Y | A X ) ≤ N Y | X , X ⊂ Y ⊆ F ∈ E ◮ Cardinality Constraints ( CC ): | R F | ≤ N F |∅ ◮ Functional Dependencies ( FD ): A X → A Y Bound Idea log | Q | ≤ maximum h ( A 1 , . . . , A n ) over all entropies h 6/22

  4. Size Bounds: Input ◮ A full conjunctive query � Q ( A [ n ] ) :- R F ( A F ) F ∈E ◮ A [ n ] = { A 1 , . . . , A n } is the full set of attributes ◮ A F = { A f | f ∈ F ⊆ [ n ] } is a subset ◮ Degree Constraints ( DC ): deg F ( A Y | A X ) ≤ N Y | X , X ⊂ Y ⊆ F ∈ E ◮ Cardinality Constraints ( CC ): | R F | ≤ N F |∅ ◮ Functional Dependencies ( FD ): A X → A Y Bound Idea log | Q | ≤ maximum h ( A 1 , . . . , A n ) over all entropies h such that h satisfies degree constraints of Q 6/22

  5. Size Bounds: Preliminaries ◮ HDC is the set of functions h : 2 [ n ] → R + satisfying the degree constraints HDC def � � = h | h ( Y | X ) ≤ log N Y | X , ∀ ( X, Y, N Y | X ) 7/22

  6. Size Bounds: Preliminaries ◮ HDC is the set of functions h : 2 [ n ] → R + satisfying the degree constraints HDC def � � = h | h ( Y | X ) ≤ log N Y | X , ∀ ( X, Y, N Y | X ) ◮ Γ ∗ n is the set of entropic functions 7/22

  7. Size Bounds: Preliminaries ◮ HDC is the set of functions h : 2 [ n ] → R + satisfying the degree constraints HDC def � � = h | h ( Y | X ) ≤ log N Y | X , ∀ ( X, Y, N Y | X ) ◮ Γ ∗ n is the set of entropic functions ∗ n is the topological closure of Γ ∗ ◮ Γ n 7/22

  8. Size Bounds: Preliminaries ◮ HDC is the set of functions h : 2 [ n ] → R + satisfying the degree constraints HDC def � � = h | h ( Y | X ) ≤ log N Y | X , ∀ ( X, Y, N Y | X ) ◮ Γ ∗ n is the set of entropic functions ∗ n is the topological closure of Γ ∗ ◮ Γ n ◮ Γ n is the set of polymatroids , i.e. functions h : 2 [ n ] → R + satisfying h ( X ∪ Y ) + h ( X ∩ Y ) ≤ h ( X ) + h ( Y ) , X, Y ⊆ [ n ] (submodularity) h ( X ) ≤ h ( Y ) , X ⊆ Y ⊆ [ n ] (monotonicity) h ( ∅ ) = 0 (strictness) 7/22

  9. Size Bounds: Preliminaries ◮ HDC is the set of functions h : 2 [ n ] → R + satisfying the degree constraints HDC def � � = h | h ( Y | X ) ≤ log N Y | X , ∀ ( X, Y, N Y | X ) ◮ Γ ∗ n is the set of entropic functions ∗ n is the topological closure of Γ ∗ ◮ Γ n ◮ Γ n is the set of polymatroids , i.e. functions h : 2 [ n ] → R + satisfying h ( X ∪ Y ) + h ( X ∩ Y ) ≤ h ( X ) + h ( Y ) , X, Y ⊆ [ n ] (submodularity) h ( X ) ≤ h ( Y ) , X ⊆ Y ⊆ [ n ] (monotonicity) h ( ∅ ) = 0 (strictness) ∗ Γ ∗ ⊂ Γ ⊂ Γ n n n ���� ���� ���� topological closure of Γ ∗ polymatroids entropic functions n 7/22

  10. Size Bounds for Full Conjunctive Queries ∗ Γ ∗ ⊂ Γ ⊂ Γ n n n ���� ���� ���� topological closure of Γ ∗ polymatroids entropic functions n 8/22

  11. Size Bounds for Full Conjunctive Queries ∗ Γ ∗ ⊂ Γ ⊂ Γ n n n ���� ���� ���� topological closure of Γ ∗ polymatroids entropic functions n Bound Entropic Bound Polymatroid Bound log | Q | ≤ log | Q | ≤ max n ∩ HDC h ([ n ]) h ∈ Γ n ∩ HDC h ([ n ]) max Definition h ∈ Γ ∗ 8/22

  12. Size Bounds for Full Conjunctive Queries ∗ Γ ∗ ⊂ Γ ⊂ Γ n n n ���� ���� ���� topological closure of Γ ∗ polymatroids entropic functions n Bound Entropic Bound Polymatroid Bound log | Q | ≤ log | Q | ≤ max n ∩ HDC h ([ n ]) h ∈ Γ n ∩ HDC h ([ n ]) max Definition h ∈ Γ ∗ AGM bound (Tight) AGM bound (Tight) CC only [Atserias et al. FOCS’08] [Atserias et al. FOCS’08] 8/22

  13. Size Bounds for Full Conjunctive Queries ∗ Γ ∗ ⊂ Γ ⊂ Γ n n n ���� ���� ���� topological closure of Γ ∗ polymatroids entropic functions n Bound Entropic Bound Polymatroid Bound log | Q | ≤ log | Q | ≤ max n ∩ HDC h ([ n ]) h ∈ Γ n ∩ HDC h ([ n ]) max Definition h ∈ Γ ∗ AGM bound (Tight) AGM bound (Tight) CC only [Atserias et al. FOCS’08] [Atserias et al. FOCS’08] Entropic Bound for FD Polymatroid Bound for FD CC + FD only [Gottlob et al. JACM’12] [Gottlob et al. JACM’12] (Tight [Gogacz et al. ICDT’17] ) (Not tight [Our work] ) 8/22

  14. Size Bounds for Full Conjunctive Queries ∗ Γ ∗ ⊂ Γ ⊂ Γ n n n ���� ���� ���� topological closure of Γ ∗ polymatroids entropic functions n Bound Entropic Bound Polymatroid Bound log | Q | ≤ log | Q | ≤ max n ∩ HDC h ([ n ]) h ∈ Γ n ∩ HDC h ([ n ]) max Definition h ∈ Γ ∗ AGM bound (Tight) AGM bound (Tight) CC only [Atserias et al. FOCS’08] [Atserias et al. FOCS’08] Entropic Bound for FD Polymatroid Bound for FD CC + FD only [Gottlob et al. JACM’12] [Gottlob et al. JACM’12] (Tight [Gogacz et al. ICDT’17] ) (Not tight [Our work] ) Entropic Bound for DC Polymatroid Bound for DC DC (Tight [Our work] ) (Not tight [Our work] ) 8/22

  15. Table of Contents Size Bounds for Full Conjunctive Queries Size Bounds for Disjunctive Datalog Algorithms for Disjunctive Datalog Algorithms for Conjunctive Queries 9/22

  16. Disjunctive Datalog � � P : T B ( A B ) :- R F ( A F ) B ∈B F ∈E 10/22

  17. Disjunctive Datalog A 2 R 12 R 23 � � A 1 A 3 P : T B ( A B ) :- R F ( A F ) R 41 R 34 B ∈B F ∈E A 4 P : T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . 10/22

  18. Disjunctive Datalog A 2 R 12 R 23 � � A 1 A 3 P : T B ( A B ) :- R F ( A F ) R 41 R 34 B ∈B F ∈E A 4 P : T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . 10/22

  19. Disjunctive Datalog A 2 R 12 R 23 � � A 1 A 3 P : T B ( A B ) :- R F ( A F ) R 41 R 34 B ∈B F ∈E A 4 P : T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . 10/22

  20. Disjunctive Datalog A 2 R 12 R 23 � � A 1 A 3 P : T B ( A B ) :- R F ( A F ) R 41 R 34 B ∈B F ∈E A 4 P : T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . A 1 A 2 A 2 A 3 A 3 A 4 A 4 A 1 a 1 1 c c 3 3 b b 1 1 d d 4 4 a b 2 2 c d 5 4 b 10/22

  21. Disjunctive Datalog A 2 R 12 R 23 � � A 1 A 3 P : T B ( A B ) :- R F ( A F ) R 41 R 34 B ∈B F ∈E A 4 P : T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . A 1 A 2 A 2 A 3 A 3 A 4 A 4 A 1 a 1 1 c c 3 3 b b 1 1 d d 4 4 a b 2 2 c d 5 4 b A 1 A 2 A 3 A 4 a 1 d 4 b 1 c 3 b 1 d 4 b 2 c 3 10/22

  22. Disjunctive Datalog A 2 R 12 R 23 � � A 1 A 3 P : T B ( A B ) :- R F ( A F ) R 41 R 34 B ∈B F ∈E A 4 P : T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . A 1 A 2 A 2 A 3 A 3 A 4 A 4 A 1 a 1 1 c c 3 3 b b 1 1 d d 4 4 a b 2 2 c d 5 4 b A 1 A 2 A 3 A 4 A 1 A 2 A 3 A 2 A 3 A 4 a 1 d 4 b 1 c 1 d 4 b 1 c 3 b 2 c 2 c 3 b 1 d 4 2 d 4 b 2 c 3 10/22

  23. Disjunctive Datalog A 2 R 12 R 23 � � A 1 A 3 P : T B ( A B ) :- R F ( A F ) R 41 R 34 B ∈B F ∈E A 4 P : T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . A 1 A 2 A 2 A 3 A 3 A 4 A 4 A 1 a 1 1 c c 3 3 b b 1 1 d d 4 4 a b 2 2 c d 5 4 b A 1 A 2 A 3 A 4 A 1 A 2 A 3 A 2 A 3 A 4 a 1 d 4 b 1 c 1 d 4 b 1 c 3 b 2 c 2 c 3 b 1 d 4 2 d 4 b 2 c 3 10/22

  24. Disjunctive Datalog A 2 R 12 R 23 � � A 1 A 3 P : T B ( A B ) :- R F ( A F ) R 41 R 34 B ∈B F ∈E A 4 P : T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . A 1 A 2 A 2 A 3 A 3 A 4 A 4 A 1 a 1 1 c c 3 3 b b 1 1 d d 4 4 a b 2 2 c d 5 4 b A 1 A 2 A 3 A 4 A 1 A 2 A 3 A 2 A 3 A 4 a 1 d 4 b 1 c 1 d 4 b 1 c 3 b 2 c 2 c 3 b 1 d 4 2 d 4 b 2 c 3 Model size is max( | T 123 | , | T 234 | ) = 3 10/22

  25. Disjunctive Datalog A 2 R 12 R 23 � � A 1 A 3 P : T B ( A B ) :- R F ( A F ) R 41 R 34 B ∈B F ∈E A 4 P : T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . A 1 A 2 A 2 A 3 A 3 A 4 A 4 A 1 a 1 1 c c 3 3 b b 1 1 d d 4 4 a b 2 2 c d 5 4 b A 1 A 2 A 3 A 4 A 1 A 2 A 3 A 2 A 3 A 4 a 1 d 4 b 1 c 1 d 4 b 1 c 3 b 2 c 2 c 3 b 1 d 4 2 d 4 b 2 c 3 Output size is the minimum over all models 10/22

  26. Disjunctive Datalog A 2 R 12 R 23 � � A 1 A 3 P : T B ( A B ) :- R F ( A F ) R 41 R 34 B ∈B F ∈E A 4 P : T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . A 1 A 2 A 2 A 3 A 3 A 4 A 4 A 1 a 1 1 c c 3 3 b b 1 1 d d 4 4 a b 2 2 c d 5 4 b A 1 A 2 A 3 A 4 A 1 A 2 A 3 A 2 A 3 A 4 a 1 d 4 b 1 c 1 d 4 b 1 c 3 b 2 c b 1 d 4 A minimum-sized model of size 2 b 2 c 3 10/22

  27. Disjunctive Datalog A 2 R 12 R 23 � � A 1 A 3 P : T B ( A B ) :- R F ( A F ) R 41 R 34 B ∈B F ∈E A 4 P : T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . A 1 A 2 A 2 A 3 A 3 A 4 A 4 A 1 a 1 1 c c 3 3 b b 1 1 d d 4 4 a b 2 2 c d 5 4 b A 1 A 2 A 3 A 4 A 1 A 2 A 3 A 2 A 3 A 4 a 1 d 4 b 1 c 1 d 4 b 1 c 3 b 2 c b 1 d 4 A minimum-sized model of size 2 b 2 c 3 ⇒ Output size is 2 10/22

  28. Disjunctive Datalog: Output Size � � P : T B ( A B ) :- R F ( A F ) B ∈B F ∈E | P ( D ) | def = min = P max B ∈B | T B | T : T | 11/22

  29. Disjunctive Datalog: Output Size � � P : T B ( A B ) :- R F ( A F ) A 2 R 12 R 23 B ∈B F ∈E A 1 A 3 | P ( D ) | def R 41 R 34 = min = P max B ∈B | T B | T : T | A 4 11/22

  30. Disjunctive Datalog: Output Size � � P : T B ( A B ) :- R F ( A F ) A 2 R 12 R 23 B ∈B F ∈E A 1 A 3 | P ( D ) | def R 41 R 34 = min = P max B ∈B | T B | T : T | A 4 | R 12 | ≤ N, | R 23 | ≤ N, | R 34 | ≤ N, | R 41 | ≤ N. D : 11/22

  31. Disjunctive Datalog: Output Size � � P : T B ( A B ) :- R F ( A F ) A 2 R 12 R 23 B ∈B F ∈E A 1 A 3 | P ( D ) | def R 41 R 34 = min = P max B ∈B | T B | T : T | A 4 | R 12 | ≤ N, | R 23 | ≤ N, | R 34 | ≤ N, | R 41 | ≤ N. D : P : T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- ◮ R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . 11/22

  32. Disjunctive Datalog: Output Size � � P : T B ( A B ) :- R F ( A F ) A 2 R 12 R 23 B ∈B F ∈E A 1 A 3 | P ( D ) | def R 41 R 34 = min = P max B ∈B | T B | T : T | A 4 | R 12 | ≤ N, | R 23 | ≤ N, | R 34 | ≤ N, | R 41 | ≤ N. D : P : T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- ◮ R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . | P ( D ) | ≤ N 3 / 2 , for all D 11/22

  33. Disjunctive Datalog: Output Size � � P : T B ( A B ) :- R F ( A F ) A 2 R 12 R 23 B ∈B F ∈E A 1 A 3 | P ( D ) | def R 41 R 34 = min = P max B ∈B | T B | T : T | A 4 | R 12 | ≤ N, | R 23 | ≤ N, | R 34 | ≤ N, | R 41 | ≤ N. D : P : T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- ◮ R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . | P ( D ) | ≤ N 3 / 2 , for all D P ′ : T 123 ( A 1 , A 2 , A 3 ) :- ◮ R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . 11/22

  34. Disjunctive Datalog: Output Size � � P : T B ( A B ) :- R F ( A F ) A 2 R 12 R 23 B ∈B F ∈E A 1 A 3 | P ( D ) | def R 41 R 34 = min = P max B ∈B | T B | T : T | A 4 | R 12 | ≤ N, | R 23 | ≤ N, | R 34 | ≤ N, | R 41 | ≤ N. D : P : T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- ◮ R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . | P ( D ) | ≤ N 3 / 2 , for all D P ′ : T 123 ( A 1 , A 2 , A 3 ) :- ◮ R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . | P ′ ( D ) | = N 2 , for some D 11/22

  35. Disjunctive Datalog: Size Bounds � � P : T B ( A B ) :- R F ( A F ) B ∈B F ∈E def | P ( D ) | = min = P max B ∈B | T B | T : T | 12/22

  36. Disjunctive Datalog: Size Bounds � � P : T B ( A B ) :- R F ( A F ) B ∈B F ∈E def | P ( D ) | = min = P max B ∈B | T B | T : T | Recall that: ◮ HDC is the set of functions h : 2 [ n ] → R + satisfying the degree constraints ∗ Γ ∗ ⊂ ⊂ Γ Γ n ◮ n n ���� ���� ���� polymatroids entropic functions topological closure of Γ ∗ n 12/22

  37. Disjunctive Datalog: Size Bounds � � P : T B ( A B ) :- R F ( A F ) B ∈B F ∈E def | P ( D ) | = min = P max B ∈B | T B | T : T | Recall that: ◮ HDC is the set of functions h : 2 [ n ] → R + satisfying the degree constraints ∗ Γ ∗ ⊂ ⊂ Γ Γ n ◮ n n ���� ���� ���� polymatroids entropic functions topological closure of Γ ∗ n Theorem log | P ( D ) | 12/22

  38. Disjunctive Datalog: Size Bounds � � P : T B ( A B ) :- R F ( A F ) B ∈B F ∈E def | P ( D ) | = min = P max B ∈B | T B | T : T | Recall that: ◮ HDC is the set of functions h : 2 [ n ] → R + satisfying the degree constraints ∗ Γ ∗ ⊂ ⊂ Γ Γ n ◮ n n ���� ���� ���� polymatroids entropic functions topological closure of Γ ∗ n Theorem log | P ( D ) | ≤ max min B ∈B h ( B ) ∗ h ∈ Γ n ∩ HDC � �� � entropic bound 12/22

  39. Disjunctive Datalog: Size Bounds � � P : T B ( A B ) :- R F ( A F ) B ∈B F ∈E def | P ( D ) | = min = P max B ∈B | T B | T : T | Recall that: ◮ HDC is the set of functions h : 2 [ n ] → R + satisfying the degree constraints ∗ Γ ∗ ⊂ ⊂ Γ Γ n ◮ n n ���� ���� ���� polymatroids entropic functions topological closure of Γ ∗ n Theorem log | P ( D ) | ≤ max B ∈B h ( B ) min ≤ max B ∈B h ( B ) min h ∈ Γ n ∩ HDC ∗ h ∈ Γ n ∩ HDC � �� � � �� � entropic bound polymatroid bound 12/22

  40. Disjunctive Datalog: Size Bounds � � P : T B ( A B ) :- R F ( A F ) B ∈B F ∈E def | P ( D ) | = min = P max B ∈B | T B | T : T | Recall that: ◮ HDC is the set of functions h : 2 [ n ] → R + satisfying the degree constraints ∗ Γ ∗ ⊂ ⊂ Γ Γ n ◮ n n ���� ���� ���� polymatroids entropic functions topological closure of Γ ∗ n Theorem log | P ( D ) | ≤ max min B ∈B h ( B ) ≤ max B ∈B h ( B ) min h ∈ Γ n ∩ HDC ∗ h ∈ Γ n ∩ HDC � �� � � �� � entropic bound polymatroid bound (asymptotically tight!) 12/22

  41. Table of Contents Size Bounds for Full Conjunctive Queries Size Bounds for Disjunctive Datalog Algorithms for Disjunctive Datalog Algorithms for Conjunctive Queries 13/22

  42. PANDA ( P roof- A ssisted e N tropic D egree- A ware) ◮ An algorithm for disjunctive datalog 14/22

  43. PANDA ( P roof- A ssisted e N tropic D egree- A ware) ◮ An algorithm for disjunctive datalog ◮ computes a model 14/22

  44. PANDA ( P roof- A ssisted e N tropic D egree- A ware) ◮ An algorithm for disjunctive datalog ◮ computes a model ◮ within the polymatroid bound: 14/22

  45. PANDA ( P roof- A ssisted e N tropic D egree- A ware) ◮ An algorithm for disjunctive datalog ◮ computes a model ◮ within the polymatroid bound: ◮ the worst-case size of the minimum model . 14/22

  46. PANDA ( P roof- A ssisted e N tropic D egree- A ware) ◮ An algorithm for disjunctive datalog ◮ computes a model ◮ within the polymatroid bound: ◮ the worst-case size of the minimum model . ◮ Outline 14/22

  47. PANDA ( P roof- A ssisted e N tropic D egree- A ware) ◮ An algorithm for disjunctive datalog ◮ computes a model ◮ within the polymatroid bound: ◮ the worst-case size of the minimum model . ◮ Outline ◮ Construct a Proof Sequence for the bound. 14/22

  48. PANDA ( P roof- A ssisted e N tropic D egree- A ware) ◮ An algorithm for disjunctive datalog ◮ computes a model ◮ within the polymatroid bound: ◮ the worst-case size of the minimum model . ◮ Outline ◮ Construct a Proof Sequence for the bound. ◮ Interpret each proof step as an algorithmic step. 14/22

  49. PANDA ◮ Polymatroid bound: h ∈ Γ n ∩ HDC min max B ∈B h ( B ) 15/22

  50. PANDA ◮ Polymatroid bound: h ∈ Γ n ∩ HDC min max B ∈B h ( B ) � h ∈ Γ n ∩ HDC min max B ∈B h ( B ) = max λ B h ( B ) ◮ h ∈ Γ n ∩ HDC B ∈B 15/22

  51. PANDA ◮ Polymatroid bound: h ∈ Γ n ∩ HDC min max B ∈B h ( B ) � h ∈ Γ n ∩ HDC min max B ∈B h ( B ) = max λ B h ( B ) ◮ h ∈ Γ n ∩ HDC B ∈B � � λ B · h ( B ) ≤ δ Y | X · h ( Y | X ) ◮ � �� � B ∈B ( X,Y,N Y | X ) ≤ log N Y | X 15/22

  52. PANDA ◮ Polymatroid bound: h ∈ Γ n ∩ HDC min max B ∈B h ( B ) � h ∈ Γ n ∩ HDC min max B ∈B h ( B ) = max λ B h ( B ) ◮ h ∈ Γ n ∩ HDC B ∈B � � λ B · h ( B ) ≤ δ Y | X · h ( Y | X ) ◮ � �� � B ∈B ( X,Y,N Y | X ) ≤ log N Y | X ◮ Proof Sequence Given X ⊆ Y : h ( X ) + h ( Y | X ) → h ( Y ) h ( Y ) → h ( X ) + h ( Y | X ) h ( Y ) → h ( X ) h ( Y | X ) → h ( Y ∪ Z | X ∪ Z ) 15/22

  53. PANDA ◮ Polymatroid bound: h ∈ Γ n ∩ HDC min max B ∈B h ( B ) � h ∈ Γ n ∩ HDC min max B ∈B h ( B ) = max λ B h ( B ) ◮ h ∈ Γ n ∩ HDC B ∈B � � λ B · h ( B ) ≤ δ Y | X · h ( Y | X ) ◮ � �� � B ∈B ( X,Y,N Y | X ) ≤ log N Y | X ◮ Proof Sequence Given X ⊆ Y : h ( X ) + h ( Y | X ) → h ( Y ) (join) h ( Y ) → h ( X ) + h ( Y | X ) (data partition) h ( Y ) → h ( X ) (projection) h ( Y | X ) → h ( Y ∪ Z | X ∪ Z ) (nothing) ◮ Algorithmic Sequence 15/22

  54. PANDA ◮ Polymatroid bound: h ∈ Γ n ∩ HDC min max B ∈B h ( B ) � h ∈ Γ n ∩ HDC min max B ∈B h ( B ) = max λ B h ( B ) ◮ h ∈ Γ n ∩ HDC B ∈B � � λ B · h ( B ) ≤ δ Y | X · h ( Y | X ) ◮ � �� � B ∈B ( X,Y,N Y | X ) ≤ log N Y | X ◮ Proof Sequence Given X ⊆ Y : h ( X ) + h ( Y | X ) → h ( Y ) (join) h ( Y ) → h ( X ) + h ( Y | X ) (data partition) h ( Y ) → h ( X ) (projection) h ( Y | X ) → h ( Y ∪ Z | X ∪ Z ) (nothing) ◮ Algorithmic Sequence Theorem PANDA solves any disjunctive datalog rule P in time within the polymatroid bound of P . 15/22

  55. PANDA : Example A 2 R 12 R 23 T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- P : A 1 A 3 R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 41 R 34 R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . A 4 16/22

  56. PANDA : Example A 2 R 12 R 23 T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- P : A 1 A 3 R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 41 R 34 R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . A 4 | R 12 | , | R 23 | , | R 34 | , | R 41 | ≤ N 16/22

  57. PANDA : Example A 2 R 12 R 23 T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- P : A 1 A 3 R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 41 R 34 R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . A 4 | P | ≤ N 3 / 2 | R 12 | , | R 23 | , | R 34 | , | R 41 | ≤ N ⇒ 16/22

  58. PANDA : Example A 2 R 12 R 23 T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- P : A 1 A 3 R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 41 R 34 R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . A 4 | P | ≤ N 3 / 2 | R 12 | , | R 23 | , | R 34 | , | R 41 | ≤ N ⇒ log | P | ≤ min( h ( A 1 A 2 A 3 ) , h ( A 2 A 3 A 4 )) 16/22

  59. PANDA : Example A 2 R 12 R 23 T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- P : A 1 A 3 R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 41 R 34 R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . A 4 | P | ≤ N 3 / 2 | R 12 | , | R 23 | , | R 34 | , | R 41 | ≤ N ⇒ 1 � � log | P | ≤ min( h ( A 1 A 2 A 3 ) , h ( A 2 A 3 A 4 )) ≤ h ( A 1 A 2 A 3 ) + h ( A 2 A 3 A 4 ) 2 16/22

  60. PANDA : Example A 2 R 12 R 23 T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- P : A 1 A 3 R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 41 R 34 R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . A 4 | P | ≤ N 3 / 2 | R 12 | , | R 23 | , | R 34 | , | R 41 | ≤ N ⇒ 1 � � log | P | ≤ min( h ( A 1 A 2 A 3 ) , h ( A 2 A 3 A 4 )) ≤ h ( A 1 A 2 A 3 ) + h ( A 2 A 3 A 4 ) 2 1 � � ≤ h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 3 A 4 ) 2 16/22

  61. PANDA : Example A 2 R 12 R 23 T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- P : A 1 A 3 R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 41 R 34 R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . A 4 | P | ≤ N 3 / 2 | R 12 | , | R 23 | , | R 34 | , | R 41 | ≤ N ⇒ 1 � � log | P | ≤ min( h ( A 1 A 2 A 3 ) , h ( A 2 A 3 A 4 )) ≤ h ( A 1 A 2 A 3 ) + h ( A 2 A 3 A 4 ) 2 1 3 � � ≤ h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 3 A 4 ) ≤ 2 log N 2 16/22

  62. PANDA : Example A 2 R 12 R 23 T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- P : A 1 A 3 R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 41 R 34 R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . A 4 | P | ≤ N 3 / 2 | R 12 | , | R 23 | , | R 34 | , | R 41 | ≤ N ⇒ 1 � � log | P | ≤ min( h ( A 1 A 2 A 3 ) , h ( A 2 A 3 A 4 )) ≤ h ( A 1 A 2 A 3 ) + h ( A 2 A 3 A 4 ) 2 1 3 � � ≤ h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 3 A 4 ) ≤ 2 log N 2 h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 3 A 4 ) 16/22

  63. PANDA : Example A 2 R 12 R 23 T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- P : A 1 A 3 R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 41 R 34 R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . A 4 | P | ≤ N 3 / 2 | R 12 | , | R 23 | , | R 34 | , | R 41 | ≤ N ⇒ 1 � � log | P | ≤ min( h ( A 1 A 2 A 3 ) , h ( A 2 A 3 A 4 )) ≤ h ( A 1 A 2 A 3 ) + h ( A 2 A 3 A 4 ) 2 1 3 � � ≤ h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 3 A 4 ) ≤ 2 log N 2 � � h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 3 A 4 ) → h ( A 3 A 4 ) → h ( A 4 | A 3 ) + h ( A 3 ) 16/22

  64. PANDA : Example A 2 R 12 R 23 T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- P : A 1 A 3 R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 41 R 34 R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . A 4 | P | ≤ N 3 / 2 | R 12 | , | R 23 | , | R 34 | , | R 41 | ≤ N ⇒ 1 � � log | P | ≤ min( h ( A 1 A 2 A 3 ) , h ( A 2 A 3 A 4 )) ≤ h ( A 1 A 2 A 3 ) + h ( A 2 A 3 A 4 ) 2 1 3 � � ≤ h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 3 A 4 ) ≤ 2 log N 2 � � h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 3 A 4 ) → h ( A 3 A 4 ) → h ( A 4 | A 3 ) + h ( A 3 ) h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 4 | A 3 ) + h ( A 3 ) 16/22

  65. PANDA : Example A 2 R 12 R 23 T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- P : A 1 A 3 R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 41 R 34 R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . A 4 | P | ≤ N 3 / 2 | R 12 | , | R 23 | , | R 34 | , | R 41 | ≤ N ⇒ 1 � � log | P | ≤ min( h ( A 1 A 2 A 3 ) , h ( A 2 A 3 A 4 )) ≤ h ( A 1 A 2 A 3 ) + h ( A 2 A 3 A 4 ) 2 1 3 � � ≤ h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 3 A 4 ) ≤ 2 log N 2 � � h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 3 A 4 ) → h ( A 3 A 4 ) → h ( A 4 | A 3 ) + h ( A 3 ) � � h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 4 | A 3 ) + h ( A 3 ) → h ( A 4 | A 3 ) → h ( A 4 | A 2 A 3 ) 16/22

  66. PANDA : Example A 2 R 12 R 23 T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- P : A 1 A 3 R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 41 R 34 R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . A 4 | P | ≤ N 3 / 2 | R 12 | , | R 23 | , | R 34 | , | R 41 | ≤ N ⇒ 1 � � log | P | ≤ min( h ( A 1 A 2 A 3 ) , h ( A 2 A 3 A 4 )) ≤ h ( A 1 A 2 A 3 ) + h ( A 2 A 3 A 4 ) 2 1 3 � � ≤ h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 3 A 4 ) ≤ 2 log N 2 � � h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 3 A 4 ) → h ( A 3 A 4 ) → h ( A 4 | A 3 ) + h ( A 3 ) � � h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 4 | A 3 ) + h ( A 3 ) → h ( A 4 | A 3 ) → h ( A 4 | A 2 A 3 ) h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 4 | A 2 A 3 ) + h ( A 3 ) 16/22

  67. PANDA : Example A 2 R 12 R 23 T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- P : A 1 A 3 R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 41 R 34 R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . A 4 | P | ≤ N 3 / 2 | R 12 | , | R 23 | , | R 34 | , | R 41 | ≤ N ⇒ 1 � � log | P | ≤ min( h ( A 1 A 2 A 3 ) , h ( A 2 A 3 A 4 )) ≤ h ( A 1 A 2 A 3 ) + h ( A 2 A 3 A 4 ) 2 1 3 � � ≤ h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 3 A 4 ) ≤ 2 log N 2 � � h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 3 A 4 ) → h ( A 3 A 4 ) → h ( A 4 | A 3 ) + h ( A 3 ) � � h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 4 | A 3 ) + h ( A 3 ) → h ( A 4 | A 3 ) → h ( A 4 | A 2 A 3 ) � � h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 4 | A 2 A 3 ) + h ( A 3 ) → h ( A 2 A 3 ) + h ( A 4 | A 2 A 3 ) → h ( A 2 A 3 A 4 ) 16/22

  68. PANDA : Example A 2 R 12 R 23 T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- P : A 1 A 3 R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 41 R 34 R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . A 4 | P | ≤ N 3 / 2 | R 12 | , | R 23 | , | R 34 | , | R 41 | ≤ N ⇒ 1 � � log | P | ≤ min( h ( A 1 A 2 A 3 ) , h ( A 2 A 3 A 4 )) ≤ h ( A 1 A 2 A 3 ) + h ( A 2 A 3 A 4 ) 2 1 3 � � ≤ h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 3 A 4 ) ≤ 2 log N 2 � � h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 3 A 4 ) → h ( A 3 A 4 ) → h ( A 4 | A 3 ) + h ( A 3 ) � � h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 4 | A 3 ) + h ( A 3 ) → h ( A 4 | A 3 ) → h ( A 4 | A 2 A 3 ) � � h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 4 | A 2 A 3 ) + h ( A 3 ) → h ( A 2 A 3 ) + h ( A 4 | A 2 A 3 ) → h ( A 2 A 3 A 4 ) h ( A 1 A 2 ) + h ( A 2 A 3 A 4 ) + h ( A 3 ) 16/22

  69. PANDA : Example A 2 R 12 R 23 T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- P : A 1 A 3 R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 41 R 34 R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . A 4 | P | ≤ N 3 / 2 | R 12 | , | R 23 | , | R 34 | , | R 41 | ≤ N ⇒ 1 � � log | P | ≤ min( h ( A 1 A 2 A 3 ) , h ( A 2 A 3 A 4 )) ≤ h ( A 1 A 2 A 3 ) + h ( A 2 A 3 A 4 ) 2 1 3 � � ≤ h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 3 A 4 ) ≤ 2 log N 2 � � h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 3 A 4 ) → h ( A 3 A 4 ) → h ( A 4 | A 3 ) + h ( A 3 ) � � h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 4 | A 3 ) + h ( A 3 ) → h ( A 4 | A 3 ) → h ( A 4 | A 2 A 3 ) � � h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 4 | A 2 A 3 ) + h ( A 3 ) → h ( A 2 A 3 ) + h ( A 4 | A 2 A 3 ) → h ( A 2 A 3 A 4 ) � � h ( A 1 A 2 ) + h ( A 2 A 3 A 4 ) + h ( A 3 ) → h ( A 1 A 2 ) → h ( A 1 A 2 | A 3 ) 16/22

  70. PANDA : Example A 2 R 12 R 23 T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- P : A 1 A 3 R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 41 R 34 R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . A 4 | P | ≤ N 3 / 2 | R 12 | , | R 23 | , | R 34 | , | R 41 | ≤ N ⇒ 1 � � log | P | ≤ min( h ( A 1 A 2 A 3 ) , h ( A 2 A 3 A 4 )) ≤ h ( A 1 A 2 A 3 ) + h ( A 2 A 3 A 4 ) 2 1 3 � � ≤ h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 3 A 4 ) ≤ 2 log N 2 � � h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 3 A 4 ) → h ( A 3 A 4 ) → h ( A 4 | A 3 ) + h ( A 3 ) � � h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 4 | A 3 ) + h ( A 3 ) → h ( A 4 | A 3 ) → h ( A 4 | A 2 A 3 ) � � h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 4 | A 2 A 3 ) + h ( A 3 ) → h ( A 2 A 3 ) + h ( A 4 | A 2 A 3 ) → h ( A 2 A 3 A 4 ) � � h ( A 1 A 2 ) + h ( A 2 A 3 A 4 ) + h ( A 3 ) → h ( A 1 A 2 ) → h ( A 1 A 2 | A 3 ) h ( A 1 A 2 | A 3 ) + h ( A 2 A 3 A 4 ) + h ( A 3 ) 16/22

  71. PANDA : Example A 2 R 12 R 23 T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- P : A 1 A 3 R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 41 R 34 R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . A 4 | P | ≤ N 3 / 2 | R 12 | , | R 23 | , | R 34 | , | R 41 | ≤ N ⇒ 1 � � log | P | ≤ min( h ( A 1 A 2 A 3 ) , h ( A 2 A 3 A 4 )) ≤ h ( A 1 A 2 A 3 ) + h ( A 2 A 3 A 4 ) 2 1 3 � � ≤ h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 3 A 4 ) ≤ 2 log N 2 � � h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 3 A 4 ) → h ( A 3 A 4 ) → h ( A 4 | A 3 ) + h ( A 3 ) � � h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 4 | A 3 ) + h ( A 3 ) → h ( A 4 | A 3 ) → h ( A 4 | A 2 A 3 ) � � h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 4 | A 2 A 3 ) + h ( A 3 ) → h ( A 2 A 3 ) + h ( A 4 | A 2 A 3 ) → h ( A 2 A 3 A 4 ) � � h ( A 1 A 2 ) + h ( A 2 A 3 A 4 ) + h ( A 3 ) → h ( A 1 A 2 ) → h ( A 1 A 2 | A 3 ) � � h ( A 1 A 2 | A 3 ) + h ( A 2 A 3 A 4 ) + h ( A 3 ) → h ( A 1 A 2 | A 3 ) + h ( A 3 ) → h ( A 1 A 2 A 3 ) 16/22

  72. PANDA : Example A 2 R 12 R 23 T 123 ( A 1 , A 2 , A 3 ) ∨ T 234 ( A 2 , A 3 , A 4 ) :- P : A 1 A 3 R 12 ( A 1 , A 2 ) , R 23 ( A 2 , A 3 ) , R 41 R 34 R 34 ( A 3 , A 4 ) , R 41 ( A 4 , A 1 ) . A 4 | P | ≤ N 3 / 2 | R 12 | , | R 23 | , | R 34 | , | R 41 | ≤ N ⇒ 1 � � log | P | ≤ min( h ( A 1 A 2 A 3 ) , h ( A 2 A 3 A 4 )) ≤ h ( A 1 A 2 A 3 ) + h ( A 2 A 3 A 4 ) 2 1 3 � � ≤ h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 3 A 4 ) ≤ 2 log N 2 � � h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 3 A 4 ) → h ( A 3 A 4 ) → h ( A 4 | A 3 ) + h ( A 3 ) � � h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 4 | A 3 ) + h ( A 3 ) → h ( A 4 | A 3 ) → h ( A 4 | A 2 A 3 ) � � h ( A 1 A 2 ) + h ( A 2 A 3 ) + h ( A 4 | A 2 A 3 ) + h ( A 3 ) → h ( A 2 A 3 ) + h ( A 4 | A 2 A 3 ) → h ( A 2 A 3 A 4 ) � � h ( A 1 A 2 ) + h ( A 2 A 3 A 4 ) + h ( A 3 ) → h ( A 1 A 2 ) → h ( A 1 A 2 | A 3 ) � � h ( A 1 A 2 | A 3 ) + h ( A 2 A 3 A 4 ) + h ( A 3 ) → h ( A 1 A 2 | A 3 ) + h ( A 3 ) → h ( A 1 A 2 A 3 ) h ( A 1 A 2 A 3 ) + h ( A 2 A 3 A 4 ) 16/22

Download Presentation
Download Policy: The content available on the website is offered to you 'AS IS' for your personal information and use only. It cannot be commercialized, licensed, or distributed on other websites without prior consent from the author. To download a presentation, simply click this link. If you encounter any difficulties during the download process, it's possible that the publisher has removed the file from their server.

Recommend

Shannon-type Inequalities, Submodular Width, and Disjunctive Datalog Hung Q. Ngo (Stealth Mode)

Shannon-type Inequalities, Submodular Width, and Disjunctive Datalog Hung Q. Ngo (Stealth Mode)

Shannon-type Inequalities, Submodular Width, and Disjunctive Datalog Hung Q. Ngo (Stealth Mode) With Mahmoud Abo Khamis and Dan Suciu @ PODS 2017 Table of Contents Connecting the Dots Output Size Bounds and Information Theory Shannon-flow

2.22k views • 190 slides
( ) Outline Submodular

( ) Outline Submodular

( ) Outline Submodular Functions Examples Discrete Convexity Minimizing Submodular Functions Symmetric Submodular Functions Maximizing Submodular Functions

567 views • 45 slides
MELODI M achin E L earning, O ptimization, & D ata I nterpretation @ UW Iyer & Bilmes,

MELODI M achin E L earning, O ptimization, & D ata I nterpretation @ UW Iyer & Bilmes,

Submodular Functions Problem Formulation Algorithmic Framework Empirical Results Submodular Optimization with Submodular Cover and Submodular Knapsack Constraints (SCSC/ SCSK) Rishabh Iyer Jeff Bilmes University of Washington, Seattle

946 views • 94 slides
The duality between Poincar e type inequalities on Hamming cube and square function

The duality between Poincar e type inequalities on Hamming cube and square function

The duality between Poincar e type inequalities on Hamming cube and square function inequalities on dyadic lattice: How to use tree to climb on hypercube based on works with Paata Ivanisvili Harmonic Analysis and applications to PDE and GMT;

851 views • 57 slides
Minimizing Submodular Functions Satoru Iwata (RIMS, Kyoto University) Outline Submodular

Minimizing Submodular Functions Satoru Iwata (RIMS, Kyoto University) Outline Submodular

Minimizing Submodular Functions Satoru Iwata (RIMS, Kyoto University) Outline Submodular Functions Examples Discrete Convexity Submodular Function Minimization Min-Max Theorem Combinatorial Algorithms Applications

625 views • 35 slides
Fast and Private Submodular and k- Submodular Functions Maximization with Matroid Constraints

Fast and Private Submodular and k- Submodular Functions Maximization with Matroid Constraints

ICML | 2020 Thirty-seventh International Conference on Machine Learning Fast and Private Submodular and k- Submodular Functions Maximization with Matroid Constraints Akbar Rafiey Yuichi Yoshida 1 Core massage What is the problem?

511 views • 22 slides
Sharp Bernstein and Hoeffding type inequalities for regenerative Markov chains Gabriela Cio

Sharp Bernstein and Hoeffding type inequalities for regenerative Markov chains Gabriela Cio

Introduction Maximal inequalities under uniform entropy Bernstein type maximal inequality for Harris recurrent Markov chains Sharp Bernstein and Hoeffding type inequalities for regenerative Markov chains Gabriela Cio lek LTCI, T el

432 views • 24 slides
Fast Semi-differential based Submodular Function Optimization Rishabh Iyer 1 Stefanie Jegelka 2

Fast Semi-differential based Submodular Function Optimization Rishabh Iyer 1 Stefanie Jegelka 2

Background Submodular Semigradients Submodular Minimization Submodular Maximization Conclusion Fast Semi-differential based Submodular Function Optimization Rishabh Iyer 1 Stefanie Jegelka 2 Jeff Bilmes 1 1 University of Washington, Seattle 2

1.24k views • 81 slides
The Muckenhoupt-type estimations for the best constants in multidimensional modular inequalities

The Muckenhoupt-type estimations for the best constants in multidimensional modular inequalities

The Muckenhoupt-type estimations for the best constants in multidimensional modular inequalities over spherical cones Chang-Pao Chen Department of Applied Mathematics Hsuan Chuang University Hsinchu, Taiwan 30092, R.O.C. Email:

574 views • 29 slides
Boundary Harnack inequalities for operators of p-Laplace type Kaj Nystrm (joint work with John

Boundary Harnack inequalities for operators of p-Laplace type Kaj Nystrm (joint work with John

Boundary Harnack inequalities for operators of p-Laplace type Kaj Nystrm (joint work with John Lewis) Ume University, Sweden Boundary Harnack inequalities for operators of p-Laplace type The p -Laplace operator G R n bounded domain, 1

405 views • 23 slides
CSE 341, Spring 2005 Lecture 27 More OO design Classic C struct Widget {width, height,

CSE 341, Spring 2005 Lecture 27 More OO design Classic C struct Widget {width, height,

CSE 341, Spring 2005 Lecture 27 More OO design Classic C struct Widget {width, height, type , type-specific fields } Widget[1000] data; x = 0; y = 0; for(i = 0; i < n; i++) { if(x+data[i].width > linewidth) {x = 0; y +=

409 views • 9 slides
Streaming -submodular Maximization under Noise subject to Size Constraint Lan N. Nguyen, My

Streaming -submodular Maximization under Noise subject to Size Constraint Lan N. Nguyen, My

Streaming -submodular Maximization under Noise subject to Size Constraint Lan N. Nguyen, My T. Thai University of Florida -submodular maximization s.t. size constraint -submodular function is a generalization of submodular

685 views • 29 slides
Abutilon palmeri COMMON NAME Indian Mallow HEIGHT WIDTH 4' 4' FOLIAGE TYPE WATER USE Evergreen Low

Abutilon palmeri COMMON NAME Indian Mallow HEIGHT WIDTH 4' 4' FOLIAGE TYPE WATER USE Evergreen Low

BOTANICAL NAME Abutilon palmeri COMMON NAME Indian Mallow HEIGHT WIDTH 4' 4' FOLIAGE TYPE WATER USE Evergreen Low PLANT TYPE FLOWER COLOR Shrub Golden/ Orange EXPOSURE FLOWER PERIOD Sun All Year SUBSTITUTE FOR BOTANICAL NAME Arctostaphylos

885 views • 39 slides
CS675: Convex and Combinatorial Optimization Fall 2019 Submodular Function Optimization

CS675: Convex and Combinatorial Optimization Fall 2019 Submodular Function Optimization

CS675: Convex and Combinatorial Optimization Fall 2019 Submodular Function Optimization Instructor: Shaddin Dughmi Outline Introduction to Submodular Functions 1 Unconstrained Submodular Minimization 2 Definition and Examples The Convex

2.22k views • 173 slides
On Birmans Sequence of Hardy-Rellich Type Inequalities Isaac B. Michael (joint with F.

On Birmans Sequence of Hardy-Rellich Type Inequalities Isaac B. Michael (joint with F.

On Birmans Sequence of Hardy-Rellich Type Inequalities Isaac B. Michael (joint with F. Gesztesy, L.L. Littlejohn and R. Wellman) IWOTA Conference - Chemnitz August 14-18, 2017 1 / 29 Introduction In 1961, M. S. Birman established the

664 views • 29 slides
Brunn-Minkowski type inequalities and conjectures K aroly B or oczky Alfr ed R

Brunn-Minkowski type inequalities and conjectures K aroly B or oczky Alfr ed R

Brunn-Minkowski type inequalities and conjectures K aroly B or oczky Alfr ed R enyi Institute of Mathematics and CEU Jena, September, 2019 Brunn-Minkowski inequality K , C convex bodies in R n , , > 0 K + C = {

484 views • 21 slides
ASSOCIATION BU I L D I N G B R I D G E S TO S T R E N G T H E N FA M I L I E S A N D C O M M

ASSOCIATION BU I L D I N G B R I D G E S TO S T R E N G T H E N FA M I L I E S A N D C O M M

WISCONSIN WIC ASSOCIATION BU I L D I N G B R I D G E S TO S T R E N G T H E N FA M I L I E S A N D C O M M U N I T I E S 2018 General Membership Meeting June 28, 2018 Angela Roltgen, RD, CLS WWA Chair W H AT I S T H E W I S C O N S I

372 views • 18 slides
Specification and Verification for distributed applications running on heterogeneous

Specification and Verification for distributed applications running on heterogeneous

Specification and Verification for distributed applications running on heterogeneous infrastructures E. Madelaine Oasis team INRIA -- CNRS - I3S -- Univ. of Nice Sophia-Antipolis ECNU Dept of Software Engineering October 2010 Eric

683 views • 49 slides
Healthcare Data & Digital Innovation Thursday 20 th April 2017 0900 - 1300 MSP Citylabs,

Healthcare Data & Digital Innovation Thursday 20 th April 2017 0900 - 1300 MSP Citylabs,

Manchester Connected Health Ecosystem Healthcare Data & Digital Innovation Thursday 20 th April 2017 0900 - 1300 MSP Citylabs, Oxford Road, Manchester Wifi Username: FREE_MSP_WIFI @Man_Inf @GM_AHSN @ECHAlliance #McrEcosystem European

1.24k views • 95 slides
CB-DBSCAN: A Novel Clustering Algorithm for Adjacent Clusters with Different Densities Gashin

CB-DBSCAN: A Novel Clustering Algorithm for Adjacent Clusters with Different Densities Gashin

CB-DBSCAN: A Novel Clustering Algorithm for Adjacent Clusters with Different Densities Gashin Ghazizadeh Outline Introduction Multi-density Clustering CB-DBSCAN Results CB-DBSCAN: A Novel Clustering Algorithm for Adjacent

454 views • 33 slides
Panda: Production and Distributed Analysis System Tadashi Maeno (BNL) on behalf of PANDA team

Panda: Production and Distributed Analysis System Tadashi Maeno (BNL) on behalf of PANDA team

Panda: Production and Distributed Analysis System Tadashi Maeno (BNL) on behalf of PANDA team Overview PanDA Production and Distributed Analysis Designed for analysis as well as production New system developed by US ATLAS team

186 views • 18 slides
The planned Nab/abBA/PANDA spectrometer Stefan Bae ler The Spallation Neutron Source SNS in Oak

The planned Nab/abBA/PANDA spectrometer Stefan Bae ler The Spallation Neutron Source SNS in Oak

The planned Nab/abBA/PANDA spectrometer Stefan Bae ler The Spallation Neutron Source SNS in Oak Ridge, TN Linear H - accelerator Accumulator ring (buncher) Target Target Guide Hall Usage of neutrons @ SNS 7 - Engineering 11A - Powder 9

742 views • 43 slides
Lecture 3: Data II How to get it, methods to parse it, and ways to explore it. Harva vard IACS

Lecture 3: Data II How to get it, methods to parse it, and ways to explore it. Harva vard IACS

Lecture 3: Data II How to get it, methods to parse it, and ways to explore it. Harva vard IACS CS109A Pavlos Protopapas, Kevin Rader, and Chris Tanner ANNOUNCEMENTS Homework 0 isnt graded for accuracy. If your questions were

618 views • 42 slides
Feasibility Studies for Nucleon Structure Measurements with PANDA Meson 2014 Ermias ATOMSSA,

Feasibility Studies for Nucleon Structure Measurements with PANDA Meson 2014 Ermias ATOMSSA,

Feasibility Studies for Nucleon Structure Measurements with PANDA Meson 2014 Ermias ATOMSSA, Binsong MA On behalf of the PANDA Collaboration Institut de Physique Nucl eaire dOrsay June 2, 2014 Meson2014, Ermias ATOMSSA,

446 views • 17 slides

More recommend