Download presentation

Presentation is loading. Please wait.

Published byKayleigh Lasswell Modified over 2 years ago

1
Time-Space Tradeoffs in Resolution: Superpolynomial Lower Bounds for Superlinear Space Chris Beck Princeton University Joint work with Paul Beame & Russell Impagliazzo

2
SAT The satisfiability problem is a central problem in computer science, in theory and in practice. Terminology: – A Clause is a boolean formula which is an OR of variables and negations of variables. – A CNF formula is an AND of clauses. Object of study for this talk: CNF-SAT: Given a CNF formula, is it satisfiable?

3
Resolution Proof System Lines are clauses, one simple proof step Proof is a sequence of clauses each of which is – an original clause or – follows from previous ones via resolution step a CNF is UNSAT iff can derive empty clause ⊥

4
Proof DAG General resolution: Arbitrary DAG Tree-like resolution: DAG is a tree

5
SAT Solvers Well-known connection between Resolution and SAT solvers based on Backtracking These algorithms are very powerful – sometimes can quickly handle CNF’s with millions of variables. On UNSAT formulas, computation history yields a Resolution proof. – Tree-like Resolution ≈ DPLL algorithm – General Resolution ≿ DPLL + “Clause Learning” Best current SAT solvers use this approach

6
SAT Solvers

7
DPLL requires very little memory Clause learning adds a new clause to the input CNF every time the search backtracks – Uses lots of memory to try to beat DPLL. – In practice, must use heuristics to guess which clauses are “important” and store only those. Hard to do well! Memory becomes a bottleneck. Question: Is this inherent? Or can the right heuristics avoid the memory bottleneck?

8
Proof Complexity & Sat Solvers Proof Size ≤ Time for Ideal SAT Solver Proof Space ≤ Memory for Ideal SAT Solver Many explicit hard UNSAT examples known with exponential lower bounds for Resolution Proof Size. Question: Is this also true for Proof Space?

9
Space in Resolution … Must be in memory

10
Lower Bounds on Space?

11
Size-Space Tradeoffs for Resolution

12
Size-Space Tradeoffs Informal Question: Can we formally show that memory rather than time can be a real bottleneck for resolution proofs and SAT solvers? Formal Question (Ben-Sasson): “Does there exist a such that any CNF with a refutation of size T also has a refutation of size T in space O()?” Our results: Families of formulas of size n having refutations in Time, Space n log n, but all resolution refutations have T > (n 0.58 log n /S) loglog n/logloglog n Not even close

13
Tseitin Tautologies Given an undirected graph and :→ 2, define a CSP: Boolean variables: Parity constraints: (linear equations) When has odd total parity, CSP is UNSAT.

14
Tseitin Tautologies When odd, G connected, corresponding CNF is called a Tseitin tautology. [Tseitin ‘68] Only total parity of matters Hard when G is a constant degree expander: [Urqhart 87]: Resolution size = 2 Ω() [Torán 99]: Resolution space =Ω(E) This work: Tradeoffs on × grid, ≫, and similar graphs, using isoperimetry.

15
Complexity Measure To measure progress as it occurs in the proof, want to define a complexity measure, which assigns a value to each clause in the proof. Wish list: – Input clauses have small value – Final clause has large value – Doesn’t grow quickly in any one resolution step

16
Complexity Measure for Tseitin Say an assignment to an (unsat) CNF is a critical if it violates only one constraint. For critical to Tseitin formula, “ ’s vertex”: For any Clause, define the “critical vertex set”: () ≔{ () : critical to, falsifies } ()≔ vertex of that constraint

17
Critical Set Examples Blue = 0 Red = 1 function: one odd vertex in corner For the empty assignment,Critical set is everything. In these examples, Graph is a Grid.

18
Critical Set Examples Blue = 0 Red = 1 For a clause that doesn’t cut the graph, Critical set is … still everything.

19
Critical Set Examples Blue = 0 Red = 1 For this clause, several components. Parity mismatch in only one, Upper left is critical.

20
Critical Set Examples Blue = 0 Red = 1

21
Complexity Measure Define () ≔ |()|. Then is a sub-additive complexity measure: (clause of ) = 1, (⊥) = #, () ≤ ( 1 ) + ( 2 ), when 1, 2 ⊦. Useful property: Every edge on the boundary of () is assigned by.

22
Graph for our result l n

23
A Refutation

24
A Different Refutation

25
Complexity vs. Time Consider the time ordering of any proof, and plot complexity of clauses in memory v. time Only constraints – start low, end high, and because of sub-additivity, cannot skip over any [t, 2t] window of μ-values on the way up. Time μ Input clauses ⊥

26
Medium complexity clauses Fix 0 = 4 and say that clause has complexity level iff 2 0 ≤ () <2 +1 0 Medium complexity clause: complexity level between 0 and log -1 – By choice of parameters 4 () < ||/2 Subadditivity can’t skip any level going up.

27
Complexity vs. Time Consider the time ordering of any proof, and divide time into equal epochs (fix later) Time Hi Med Low

28
Two Possibilities Consider the time ordering of any proof, and divide time into equal epochs (fix later) Either, a clause of medium complexity appears in memory for at least one of the breakpoints between epochs, Time Hi Med Low

29
Two Possibilities Time Hi Med Low

30
Don’t know how to use this directly Not much to go on with formula for original graph However, a Tseitin formula for graph G contains Tseitin formulas for all subgraphs of G – Just set edge variables to 0,1 to delete them. Refutation for G contains refutations for all subgraphs of G

31
Idea: Random Restrictions A restriction is a partial assignment of truth values to variables, simplifying formulas. a CNF, yields restricted formula | Π a proof of, yields restricted refutation Π| of | -size, space don’t increase. Idea: Choose ρ randomly, and see which of the two possibilities happens in restricted proof.

32
Medium Sets have large Boundary

34
Intuition Time Hi Med Low

35
Intuition Time divided into epochs If ( ⁄ ) is also small, we show that this 2nd scenario is also unlikely in the restricted proof... Time Hi Med Low

36
Extended Isoperimetric Inequality Lemma: For 1, 2 ⊆, both medium, 2| 1 |<| 2 |, have |( 1 ) ∪ ( 2 )| ≥ 2 2. Two medium sets of very different sizes boundary guarantee doubles. Also: medium sets of super-increasing sizes boundary guarantee goes up by factor of

37
Consequence of Extended Isoperimetric Inequality General Lemma: If we set each edge to 0,1 or leave unset each with prob=1/3, then for any fixed set of clauses, probability to end up with k distinct medium complexity levels is at most exp(− 2 )

38
A tradeoff Time Hi Med Low

39
An even better tradeoff Don’t just divide into epochs once Recursively divide proof into epochs and sub-epochs where each sub-epoch contains sub-epochs of the next smaller size Prove that if a lot of progress happens in some epoch, and the breakpoints of its sub-epochs don’t contain many different levels, then a lot of progress happens in some sub-epoch

40
An even better tradeoff a

42
Final Tradeoff If space is at most 2 2 / 2 then size blows up by a super-polynomial amount

43
Proof DAG

44
“Regular”: On every root to leaf path, no variable resolved more than once.

45
Tradeoffs for Regular Resolution Theorem : For any k, 4-CNF formulas (Tseitin formulas on long and skinny grid graphs) of size n with – Regular resolution refutations in size n k+1, Space n k. – But with Space only n k- , for any > 0, any regular resolution refutation requires size at least n log log n / log log log n.

46
Regular Resolution Can define partial information more precisely Complexity is monotonic wrt proof DAG edges. This part uses regularity assumption, simplifies arguments with complexity plot. Random Adversary selects random assignments based on proof – No random restrictions, conceptually clean and don’t lose constant factors here and there.

47
Open Questions More than quasi-polynomial separations? – For Tseitin formulas upper bound for small space is only a log n power of the unrestricted size – Candidate formulas? – Are these even possible? Other proof systems? Other cases for separating search paradigms: “dynamic programming” vs. “binary search”?

48
Thanks!

Similar presentations

OK

Walks, Paths and Circuits Walks, Paths and Circuits Sanjay Jain, Lecturer, School of Computing.

Walks, Paths and Circuits Walks, Paths and Circuits Sanjay Jain, Lecturer, School of Computing.

© 2017 SlidePlayer.com Inc.

All rights reserved.

Ads by Google

Ppt on nature and human personality Ppt on world diabetes day theme Ppt on indian politics bjp Lcos display ppt online Maths ppt on rational numbers for class 8 Ppt on index of industrial production Ppt on year of faith Ppt on manufacturing of ball bearing Ppt online application form 2016 Ppt on email etiquettes presentation