formalizing randomized matching algorithms
play

Formalizing Randomized Matching Algorithms Dai Tri Man L e and - PowerPoint PPT Presentation

Formalizing Randomized Matching Algorithms Dai Tri Man L e and Stephen Cook Department of Computer Science University of Toronto Canada LICS 2011 L e and Cook (University of Toronto) 1 / 16 Two Aspects of Proof Complexity Propositional


  1. Formalizing Randomized Matching Algorithms Dai Tri Man Lˆ e and Stephen Cook Department of Computer Science University of Toronto Canada LICS 2011 Lˆ e and Cook (University of Toronto) 1 / 16

  2. Two Aspects of Proof Complexity Propositional Proof Complexity (Pitassi’s invited talk) 1 the lengths of proofs of tautologies in various proof systems Bounded Arithmetic 2 the power of weak formal systems to prove theorems of interest in computer science (1) and (2) are related by “propositional translations” a proof in theory T � uniform short proofs in propositional system P T bounded arithmetic = uniform version of propositional proof complexity “bounded”: induction axioms are restricted to bounded formulas Lˆ e and Cook (University of Toronto) 2 / 16

  3. Two Aspects of Proof Complexity Propositional Proof Complexity (Pitassi’s invited talk) 1 the lengths of proofs of tautologies in various proof systems Bounded Arithmetic 2 the power of weak formal systems to prove theorems of interest in computer science (1) and (2) are related by “propositional translations” a proof in theory T � uniform short proofs in propositional system P T bounded arithmetic = uniform version of propositional proof complexity “bounded”: induction axioms are restricted to bounded formulas Lˆ e and Cook (University of Toronto) 2 / 16

  4. Bounded Arithmetic - Main Goals Complexity Theory Bounded Arithmetic Classify theorems according to the computational complexity of Classify problems according concepts needed to prove them. to complexity classes “Bounded Reverse Mathematics” [Cook-Nguyen ’10] Separate (or collapse) Separate (or collapse) complexity classes formal theories for various complexity classes Lˆ e and Cook (University of Toronto) 3 / 16

  5. Bounded Arithmetic - Main Goals Complexity Theory Bounded Arithmetic Classify theorems according to the computational complexity of Classify problems according concepts needed to prove them. to complexity classes “Bounded Reverse Mathematics” [Cook-Nguyen ’10] Separate (or collapse) Separate (or collapse) complexity classes formal theories for various complexity classes Lˆ e and Cook (University of Toronto) 3 / 16

  6. Feasible reasoning with VPV The VPV theory associated with complexity class P (polytime) universal theory based on Cook’s theory PV (’75) with symbols for all polytime functions and their defining axioms based on Cobham’s Theorem (’65). Induction on polytime predicates: a derived result via binary search. Proposition translation: polynomial size extended Frege proofs Lˆ e and Cook (University of Toronto) 4 / 16

  7. Feasible reasoning with VPV The VPV theory associated with complexity class P (polytime) universal theory based on Cook’s theory PV (’75) with symbols for all polytime functions and their defining axioms based on Cobham’s Theorem (’65). Induction on polytime predicates: a derived result via binary search. Proposition translation: polynomial size extended Frege proofs Proofs in VPV are feasibly constructive . Given a proof in VPV for the formula ∀ X ∃ Y ϕ ( X , Y ), where ϕ represents a polytime predicate, we can extract a polytime function F ( X ) and a correctness proof in VPV of ∀ X ϕ ( X , F ( X )). Induction is restricted to polytime “concepts”. Lˆ e and Cook (University of Toronto) 4 / 16

  8. Feasible proofs Polytime algorithms usually have feasible correctness proofs, e.g., the “augmenting-path” algorithm: finding a maximum matching the Hungarian algorithm: finding a minimum-weight matching . . . (formalized in VPV , see the full version on our websites) Lˆ e and Cook (University of Toronto) 5 / 16

  9. Feasible proofs Polytime algorithms usually have feasible correctness proofs, e.g., the “augmenting-path” algorithm: finding a maximum matching the Hungarian algorithm: finding a minimum-weight matching . . . (formalized in VPV , see the full version on our websites) Main Question How about randomized algorithms and probabilistic reasoning? “Formalizing Randomized Matching Algorithms” Lˆ e and Cook (University of Toronto) 5 / 16

  10. How about randomized algorithms? Two fundamental randomized matching algorithms RNC 2 algorithm for testing if a bipartite graph has a perfect matching 1 (Lov´ asz ’79) RNC 2 algorithm for finding a perfect matching of a bipartite graph 2 (Mulmuley-Vazirani-Vazirani ’87) Recall that: Log-Space ⊆ NC 2 ⊆ P RNC 2 ⊆ RP Important Remark The two algorithms above also work for general undirected graphs, but we only consider bipartite graphs. Lˆ e and Cook (University of Toronto) 6 / 16

  11. How about randomized algorithms? Two fundamental randomized matching algorithms RNC 2 algorithm for testing if a bipartite graph has a perfect matching 1 (Lov´ asz ’79) RNC 2 algorithm for finding a perfect matching of a bipartite graph 2 (Mulmuley-Vazirani-Vazirani ’87) Recall that: Log-Space ⊆ NC 2 ⊆ P RNC 2 ⊆ RP Important Remark The two algorithms above also work for general undirected graphs, but we only consider bipartite graphs. Lˆ e and Cook (University of Toronto) 6 / 16

  12. Lov´ asz’s Algorithm a c b Problem: Given a bipartite graph G , decide if G has a perfect matching. e d f d e f replace ones with a  1 0 1    x 11 0 x 13 distinct variables b 1 1 0 M G = x 21 x 22 0     c 0 1 1 0 x 32 x 33 Edmonds’ Theorem (provable in VPV) G has a perfect matching if and only if Det( M G ) is not identically zero. Lˆ e and Cook (University of Toronto) 7 / 16

  13. Lov´ asz’s Algorithm a c b Problem: Given a bipartite graph G , decide if G has a perfect matching. e d f d e f replace ones with a  1 0 1    x 11 0 x 13 distinct variables b 1 1 0 M G = x 21 x 22 0     c 0 1 1 0 x 32 x 33 Edmonds’ Theorem (provable in VPV) G has a perfect matching if and only if Det( M G ) is not identically zero. The usual proof is not feasible since. . . σ ∈ S n sgn( σ ) � n it uses the formula Det( M ) = � i =1 M ( i , σ ( i )), which has n ! terms. Lˆ e and Cook (University of Toronto) 7 / 16

  14. Lov´ asz’s Algorithm d e f replace ones with     a 1 0 1 x 11 0 x 13 distinct variables b 1 1 0 M G = x 21 x 22 0     c 0 1 1 0 x 32 x 33 Edmonds’ Theorem (provable in VPV) G has a perfect matching if and only if Det( M G ) is not identically zero. Lˆ e and Cook (University of Toronto) 8 / 16

  15. Lov´ asz’s Algorithm d e f replace ones with     a 1 0 1 x 11 0 x 13 distinct variables b 1 1 0 M G = x 21 x 22 0     c 0 1 1 0 x 32 x 33 Edmonds’ Theorem (provable in VPV) G has a perfect matching if and only if Det( M G ) is not identically zero. Observation: instance of the polynomial identity testing problem ) is a polynomial in n 2 variables x ij with degree at most n . Det( M n × n G Det( M G ) is called the Edmonds’ polynomial of G . Lˆ e and Cook (University of Toronto) 8 / 16

  16. Lov´ asz’s Algorithm Edmonds’ Theorem (provable in VPV) G has a perfect matching if and only if Det( M G ) is not identically zero. Observation: instance of the polynomial identity testing problem ) is a polynomial in n 2 variables x ij with degree at most n . Det( M n × n G Det( M G ) is called the Edmonds’ polynomial of G . asz’s RNC 2 Algorithm Lov´ Pick n 2 random values r ij from S = { 0 , . . . , 2 n } If Det( M G )( � r ) = 0 then YES (Det( M G ) ≡ 0) else NO. Lˆ e and Cook (University of Toronto) 8 / 16

  17. Lov´ asz’s Algorithm Edmonds’ Theorem (provable in VPV) G has a perfect matching if and only if Det( M G ) is not identically zero. Observation: instance of the polynomial identity testing problem ) is a polynomial in n 2 variables x ij with degree at most n . Det( M n × n G Det( M G ) is called the Edmonds’ polynomial of G . asz’s RNC 2 Algorithm Lov´ Pick n 2 random values r ij from S = { 0 , . . . , 2 n } If Det( M G )( � r ) = 0 then YES (Det( M G ) ≡ 0) else NO. if Det( M G ) ≡ 0, then Det( M G )( � r ) = 0 1 � � if Det( M G ) �≡ 0, then Pr � Det( M G )( � r ) � = 0 ≥ 1 / 2 2 r ∈ R S n 2 ((2) follows from the Schwartz-Zippel Lemma) Lˆ e and Cook (University of Toronto) 8 / 16

  18. Obstacle #1 - Talking about probability Given a polytime predicate A ( X , R ), = |{ R ∈ { 0 , 1 } n | A ( X , R ) }| � � Pr R ∈{ 0 , 1 } n A ( X , R ) 2 n R ∈ { 0 , 1 } n | A ( X , R ) � � The function F ( X ) := | | is in #P. #P problems are generally harder than NP problems Lˆ e and Cook (University of Toronto) 9 / 16

  19. Obstacle #1 - Talking about probability Given a polytime predicate A ( X , R ), = |{ R ∈ { 0 , 1 } n | A ( X , R ) }| � � Pr R ∈{ 0 , 1 } n A ( X , R ) 2 n R ∈ { 0 , 1 } n | A ( X , R ) � � The function F ( X ) := | | is in #P. #P problems are generally harder than NP problems Solution [Jeˇ r´ abek ’04] � � We want to show Pr R ∈{ 0 , 1 } n A ( X , R ) ≥ s / t , it suffices to show |{ R ∈ { 0 , 1 } n | A ( X , R ) }| · t ≥ 2 n · s Key idea: construct in VPV a polytime surjection G : { R ∈ { 0 , 1 } n | A ( X , R ) } × [ t ] ։ { 0 , 1 } n × [ s ] , where [ m ] := { 1 , . . . , m } . Lˆ e and Cook (University of Toronto) 9 / 16

Recommend


More recommend