on deciding mus membership with qbf
play

On Deciding MUS Membership with QBF s Janota 1 Joao Marques-Silva 1 , - PowerPoint PPT Presentation

On Deciding MUS Membership with QBF s Janota 1 Joao Marques-Silva 1 , 2 Mikol a 1 INESC-ID/IST, Lisbon, Portugal 2 CASL/CSI, University College Dublin, Ireland (INESC-ID & UCD) cmMUS 1 / 18 CNF and Unsatisfiability y , { x y ,


  1. On Deciding MUS Membership with QBF s Janota 1 Joao Marques-Silva 1 , 2 Mikol´ aˇ 1 INESC-ID/IST, Lisbon, Portugal 2 CASL/CSI, University College Dublin, Ireland (INESC-ID & UCD) cmMUS 1 / 18

  2. CNF and Unsatisfiability ¬ y , { x ∨ y , ¬ x , z } (INESC-ID & UCD) cmMUS 2 / 18

  3. CNF and Unsatisfiability ¬ y , { x ∨ y , ¬ x , z } (INESC-ID & UCD) cmMUS 2 / 18

  4. CNF and Unsatisfiability ¬ y , { x ∨ y , ¬ x , z } MUS An UNSAT set of clauses that becomes SAT by removing any clause is called m inimally u nsatisfiable s et (MUS) (INESC-ID & UCD) cmMUS 2 / 18

  5. CNF and Unsatisfiability ¬ y , { x ∨ y , ¬ x , z } MUS An UNSAT set of clauses that becomes SAT by removing any clause is called m inimally u nsatisfiable s et (MUS) MUS-Membership IN: a clause ω and a CNF φ Q: Is there an MUS ψ ⊆ φ such that ω ∈ ψ ? (INESC-ID & UCD) cmMUS 2 / 18

  6. Motivation Restoring Consistency Removing a clause that is not part of any MUS, will certainly not restore consistency. (INESC-ID & UCD) cmMUS 3 / 18

  7. Motivation Restoring Consistency Removing a clause that is not part of any MUS, will certainly not restore consistency. Product Configuration When configuring a product, some sets of its features result in an inconsistent configuration. Clearly, it is useful for the user(s) to know if a feature is relevant for the inconsistency. (INESC-ID & UCD) cmMUS 3 / 18

  8. How Hard Is It? x 1 , x 1 → z , { x 2 , x 2 → z , y 1 , y 1 → ¬ z , y 2 , y 2 → ¬ z , ω } (INESC-ID & UCD) cmMUS 4 / 18

  9. How Hard Is It? x 1 , x 1 → z , { x 2 , x 2 → z , y 1 , y 1 → ¬ z , y 2 , y 2 → ¬ z , ω } (INESC-ID & UCD) cmMUS 4 / 18

  10. How Hard Is It? x 1 , x 1 → z , { x 2 , x 2 → z , y 1 , y 1 → ¬ z , y 2 , y 2 → ¬ z , ω } (INESC-ID & UCD) cmMUS 4 / 18

  11. How Hard Is It? x 1 , x 1 → z , { x 2 , x 2 → z , y 1 , y 1 → ¬ z , y 2 , y 2 → ¬ z , ω } (INESC-ID & UCD) cmMUS 4 / 18

  12. How Hard Is It? x 1 , x 1 → z , { x 2 , x 2 → z , y 1 , y 1 → ¬ z , y 2 , y 2 → ¬ z , ω } (INESC-ID & UCD) cmMUS 4 / 18

  13. How Hard Is It? x 1 , x 1 → z , { x 2 , x 2 → z , y 1 , y 1 → ¬ z , y 2 , y 2 → ¬ z , ω } MUS-Membership is Σ P 2 -complete [Kul07] (INESC-ID & UCD) cmMUS 4 / 18

  14. Approaches to the Problem QBF 2 , ∃ , O ( n 2 ) QBF 3 , ∃ , O ( n ) MUS-membership MSS-membership QBF 2 , ∃ , O ( n ) Circ-Infer, O ( n ) [JMS11] [JGMS10] (INESC-ID & UCD) cmMUS 5 / 18

  15. Quantifying over Subsets Relaxation φ ∗ = { c ∨ r c | c ∈ φ } (INESC-ID & UCD) cmMUS 6 / 18

  16. Quantifying over Subsets Relaxation φ ∗ = { c ∨ r c | c ∈ φ } Relaxing Clauses Example φ = { x ∨ y , ¬ x , ¬ y } φ ∗ = { r 1 ∨ x ∨ y , r 2 ∨ ¬ x , r 3 ∨ ¬ y } (INESC-ID & UCD) cmMUS 6 / 18

  17. Quantifying over Subsets Relaxation φ ∗ = { c ∨ r c | c ∈ φ } Relaxing Clauses Example φ = { x ∨ y , ¬ x , ¬ y } φ ∗ = { r 1 ∨ x ∨ y , r 2 ∨ ¬ x , r 3 ∨ ¬ y } r 1 = 0 r 1 ∨ x ∨ y r 2 = 0 r 2 ∨ ¬ x r 3 = 1 r 3 ∨ ¬ y (INESC-ID & UCD) cmMUS 6 / 18

  18. Quantifying over Subsets Relaxation φ ∗ = { c ∨ r c | c ∈ φ } Relaxing Clauses Example φ = { x ∨ y , ¬ x , ¬ y } φ ∗ = { r 1 ∨ x ∨ y , r 2 ∨ ¬ x , r 3 ∨ ¬ y } r 1 = 0 r 1 ∨ x ∨ y r 2 = 0 r 2 ∨ ¬ x r 3 = 1 r 3 ∨ ¬ y (INESC-ID & UCD) cmMUS 6 / 18

  19. Modeling Elements Membership ∃ R . ¬ r ω (INESC-ID & UCD) cmMUS 7 / 18

  20. Modeling Elements Membership ∃ R . ¬ r ω Unsat ∃ R . ∀ X . ¬ φ ∗ ( R , X ) (INESC-ID & UCD) cmMUS 7 / 18

  21. Modeling Elements Membership ∃ R . ¬ r ω Unsat ∃ R . ∀ X . ¬ φ ∗ ( R , X ) Subset R = { r 1 , . . . , r n } , R ′ = { r ′ 1 , . . . , r ′ n } R < R ′ ≡ � � r i ⇒ r ′ ¬ r i ∧ r ′ i ∧ i r i ∈ R r i ∈ R (INESC-ID & UCD) cmMUS 7 / 18

  22. Na¨ ıve Approaches Schema exists ψ ⊆ φ s.t. ω ∈ ψ and ψ is unsatisfiable and forall ψ ′ � ψ is satisfiable (INESC-ID & UCD) cmMUS 8 / 18

  23. Na¨ ıve Approaches Schema exists ψ ⊆ φ s.t. ω ∈ ψ and ψ is unsatisfiable and forall ψ ′ � ψ is satisfiable 3-level quantification ∃ R . ¬ r ω ∧ ( ∀ X . ¬ φ ∗ ( R , X )) ∧ ( ∀ R ′ . ( R < R ′ ) ⇒ ∃ X ′ .φ ∗ ( R ′ , X ′ )) (INESC-ID & UCD) cmMUS 8 / 18

  24. Na¨ ıve Approaches Schema exists ψ ⊆ φ s.t. ω ∈ ψ and ψ is unsatisfiable and forall ψ ′ � ψ is satisfiable 3-level quantification ∃ R . ¬ r ω ∧ ( ∀ X . ¬ φ ∗ ( R , X )) ∧ ( ∀ R ′ . ( R < R ′ ) ⇒ ∃ X ′ .φ ∗ ( R ′ , X ′ )) 2-level quantification, O ( n 2 ) ∃ R . ¬ r ω ∧ ( ∀ X . ¬ φ ∗ ( R , X )) ∧ r ω i ∈ R ( ¬ r ω i ⇒ ∃ X ω i .φ ∗ [ r ω i / 1]( R , X ω i )) � (INESC-ID & UCD) cmMUS 8 / 18

  25. Na¨ ıve Approaches Schema exists ψ ⊆ φ s.t. ω ∈ ψ and ψ is unsatisfiable and forall ψ ′ � ψ is satisfiable 3-level quantification ∃ R . ¬ r ω ∧ ( ∀ X . ¬ φ ∗ ( R , X )) ∧ ( ∀ R ′ . ( R < R ′ ) ⇒ ∃ X ′ .φ ∗ ( R ′ , X ′ )) 2-level quantification, O ( n 2 ) ∃ R . ¬ r ω ∧ ( ∀ X . ¬ φ ∗ ( R , X )) ∧ r ω i ∈ R ( ¬ r ω i ⇒ ∃ X ω i .φ ∗ [ r ω i / 1]( R , X ω i )) � 2-level quantification, O ( n 2 ), prefix form ∃ RX ω 1 . . . ∃ X ω n ∀ X . ¬ r ω ∧ ¬ φ ∗ ( R , X ) ∧ r ω i ∈ R ( ¬ r ω i ⇒ φ ∗ [ r ω i / 1]( R , X ω i )) � (INESC-ID & UCD) cmMUS 8 / 18

  26. Approaches to the Problem QBF 2 , ∃ , O ( n 2 ) QBF 3 , ∃ , O ( n ) MUS-membership MSS-membership QBF 2 , ∃ , O ( n ) Circ-Infer, O ( n ) [JMS11] [JGMS10] (INESC-ID & UCD) cmMUS 9 / 18

  27. From MUS-Membership to MSS-Membership MSS A set of clauses ψ ⊆ φ is a Maximally Satisfiable Subset (MSS) iff ψ is satisfiable and any set ψ ′ ⊆ φ such that ψ � ψ ′ is unsatisfiable. (INESC-ID & UCD) cmMUS 10 / 18

  28. From MUS-Membership to MSS-Membership MSS A set of clauses ψ ⊆ φ is a Maximally Satisfiable Subset (MSS) iff ψ is satisfiable and any set ψ ′ ⊆ φ such that ψ � ψ ′ is unsatisfiable. MSS-membership IN: A CNF formula φ and a clause ω ∈ φ . Q: Is there an MSS ψ of φ such that ω / ∈ ψ ? (INESC-ID & UCD) cmMUS 10 / 18

  29. MUS-Membership ↔ MSS-membership A clause ω belongs to some MUS iff there is some MSS that does not contain ω . (INESC-ID & UCD) cmMUS 11 / 18

  30. MUS-Membership ↔ MSS-membership A clause ω belongs to some MUS iff there is some MSS that does not contain ω . MSS-membership to QBF ∃ R ∃ X ∀ R ′ ∀ X ′ . r ω ∧ φ ∗ ( R , X ) ∧ ( R ′ < R ⇒ ¬ φ ∗ ( R ′ , X ′ )) (INESC-ID & UCD) cmMUS 11 / 18

  31. Minimal Models A model of a formula is V-minimal iff flipping any subset of 1-values of variables from V to 0, yields a non-model. (INESC-ID & UCD) cmMUS 12 / 18

  32. Minimal Models A model of a formula is V-minimal iff flipping any subset of 1-values of variables from V to 0, yields a non-model. x ∨ ( y ∧ z ) [1,1,1] [1,1,0] [1,0,1] [0,1,1] [1,0,0] [0,1,0] [0,0,1] [0,0,0] (INESC-ID & UCD) cmMUS 12 / 18

  33. Minimal Models A model of a formula is V-minimal iff flipping any subset of 1-values of variables from V to 0, yields a non-model. x ∨ ( y ∧ z ) [1,1,1] [1,1,0] [1,0,1] [0,1,1] [1,0,0] [0,1,0] [0,0,1] [0,0,0] (INESC-ID & UCD) cmMUS 12 / 18

  34. Minimal Models A model of a formula is V-minimal iff flipping any subset of 1-values of variables from V to 0, yields a non-model. x ∨ ( y ∧ z ) [1,1,1] [0,1,1] [1,1,0] [1,0,1] [0,1,1] [1,0,0] [1,0,0] [0,1,0] [0,0,1] [0,0,0] (INESC-ID & UCD) cmMUS 12 / 18

  35. Entailment in Circumscription CircInfer IN: τ and ψ be propositional formulas Q: Does ψ hold in all minimal models of τ . τ | = min ψ (INESC-ID & UCD) cmMUS 13 / 18

  36. Entailment in Circumscription CircInfer IN: τ and ψ be propositional formulas Q: Does ψ hold in all minimal models of τ . τ | = min ψ CircInfer , complexity = min ψ is in Π P Deciding τ | 2 -complete [EG93] (INESC-ID & UCD) cmMUS 13 / 18

  37. MSSes and Minimal Models . . . r 1 x φ = { x , ¬ x , z } r 2 . . . ¬ x . . . r 3 z (INESC-ID & UCD) cmMUS 14 / 18

  38. MSSes and Minimal Models . . . r 1 x φ = { x , ¬ x , z } r 2 . . . ¬ x . . . r 3 z {} , [1,1,1] { x } , [0,1,1] {¬ x } , [1,0,1] { z } , [1,1,0] { x , ¬ x } , [0,0,1] { x , z } , [0,1,0] {¬ x , z } , [1,0,0] { x , ¬ x , z } , [0,0,0] (INESC-ID & UCD) cmMUS 14 / 18

  39. MSSes and Minimal Models . . . r 1 x φ = { x , ¬ x , z } r 2 . . . ¬ x . . . r 3 z {} , [1,1,1] { x } , [0,1,1] {¬ x } , [1,0,1] { z } , [1,1,0] { x , ¬ x } , [0,0,1] { x , z } , [0,1,0] {¬ x , z } , [1,0,0] { x , ¬ x , z } , [0,0,0] (INESC-ID & UCD) cmMUS 14 / 18

  40. MSSes and Minimal Models . . . r 1 x φ = { x , ¬ x , z } r 2 . . . ¬ x . . . r 3 z {} , [1,1,1] { x } , [0,1,1] {¬ x } , [1,0,1] { z } , [1,1,0] { x , ¬ x } , [0,0,1] { x , z } , [0,1,0] { x , z } , [0,1,0] {¬ x , z } , [1,0,0] {¬ x , z } , [1,0,0] { x , ¬ x , z } , [0,0,0] (INESC-ID & UCD) cmMUS 14 / 18

  41. From MSS-Membership to CircInfer MSSes ↔ Min. Models MSSes correspond to R -minimal models of φ ∗ ( R , X ). (INESC-ID & UCD) cmMUS 15 / 18

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


More recommend