Solution Counting Methods for Combinatorial Problems Ashish Sabharwal [ Cornell University] Based on joint work with: Carla Gomes, Willem-Jan van Hoeve,

Slides:



Advertisements
Similar presentations
Adnan Darwiche Computer Science Department UCLA
Advertisements

Sampling and Soundness: Can We Have Both? Carla Gomes, Bart Selman, Ashish Sabharwal Cornell University Jörg Hoffmann DERI Innsbruck …and I am: Frank van.
Lower Bounds for Exact Model Counting and Applications in Probabilistic Databases Paul Beame Jerry Li Sudeepa Roy Dan Suciu University of Washington.
Propositional Satisfiability (SAT) Toby Walsh Cork Constraint Computation Centre University College Cork Ireland 4c.ucc.ie/~tw/sat/
10/7/2014 Constrainedness of Search Toby Walsh NICTA and UNSW
Time-Space Tradeoffs in Resolution: Superpolynomial Lower Bounds for Superlinear Space Chris Beck Princeton University Joint work with Paul Beame & Russell.
Dana Nau: Lecture slides for Automated Planning Licensed under the Creative Commons Attribution-NonCommercial-ShareAlike License:
UIUC CS 497: Section EA Lecture #2 Reasoning in Artificial Intelligence Professor: Eyal Amir Spring Semester 2004.
Queries with Difference on Probabilistic Databases Sanjeev Khanna Sudeepa Roy Val Tannen University of Pennsylvania 1.
Counting the bits Analysis of Algorithms Will it run on a larger problem? When will it fail?
CPSC 422, Lecture 21Slide 1 Intelligent Systems (AI-2) Computer Science cpsc422, Lecture 21 Mar, 4, 2015 Slide credit: some slides adapted from Stuart.
Time-Space Tradeoffs in Resolution: Superpolynomial Lower Bounds for Superlinear Space Chris Beck Princeton University Joint work with Paul Beame & Russell.
1 Backdoor Sets in SAT Instances Ryan Williams Carnegie Mellon University Joint work in IJCAI03 with: Carla Gomes and Bart Selman Cornell University.
Junction Trees And Belief Propagation. Junction Trees: Motivation What if we want to compute all marginals, not just one? Doing variable elimination for.
Counting Solution Clusters Divide-and-Conquer Recursively:  Arbitrarily pick a variable, say x, of formula F  Count how many clusters contain solutions.
MPE, MAP AND APPROXIMATIONS Lecture 10: Statistical Methods in AI/ML Vibhav Gogate The University of Texas at Dallas Readings: AD Chapter 10.
Approximate Counting via Correlation Decay Pinyan Lu Microsoft Research.
Leveraging Belief Propagation, Backtrack Search, and Statistics for Model Counting Lukas Kroc, Ashish Sabharwal, Bart Selman Cornell University May 23,
CP Formal Models of Heavy-Tailed Behavior in Combinatorial Search Hubie Chen, Carla P. Gomes, and Bart Selman
1 Boolean Satisfiability in Electronic Design Automation (EDA ) By Kunal P. Ganeshpure.
1 Towards Efficient Sampling: Exploiting Random Walk Strategy Wei Wei, Jordan Erenrich, and Bart Selman.
Model Counting: A New Strategy for Obtaining Good Bounds Carla P. Gomes, Ashish Sabharwal, Bart Selman Cornell University AAAI Conference, 2006 Boston,
Combinatorial Problems II: Counting and Sampling Solutions Ashish Sabharwal Cornell University March 4, nd Asian-Pacific School on Statistical Physics.
1 Sampling, Counting, and Probabilistic Inference Wei joint work with Bart Selman.
AAAI00 Austin, Texas Generating Satisfiable Problem Instances Dimitris Achlioptas Microsoft Carla P. Gomes Cornell University Henry Kautz University of.
Analysis of Algorithms CS 477/677
Short XORs for Model Counting: From Theory to Practice Carla P. Gomes, Joerg Hoffmann, Ashish Sabharwal, Bart Selman Cornell University & Univ. of Innsbruck.
1 Backdoors To Typical Case Complexity Ryan Williams Carnegie Mellon University Joint work with: Carla Gomes and Bart Selman Cornell University.
Carla P. Gomes CS4700 CS 4700: Foundations of Artificial Intelligence Carla P. Gomes Module: Instance Hardness and Phase Transitions.
1 CS 4700: Foundations of Artificial Intelligence Carla P. Gomes Module: Satisfiability (Reading R&N: Chapter 7)
ENGG3190 Logic Synthesis “Boolean Satisfiability” Winter 2014 S. Areibi School of Engineering University of Guelph.
1 Combinatorial Problems in Cooperative Control: Complexity and Scalability Carla Gomes and Bart Selman Cornell University Muri Meeting March 2002.
1 Message Passing and Local Heuristics as Decimation Strategies for Satisfiability Lukas Kroc, Ashish Sabharwal, Bart Selman (presented by Sebastian Brand)
Sampling Combinatorial Space Using Biased Random Walks Jordan Erenrich, Wei Wei and Bart Selman Dept. of Computer Science Cornell University.
Relaxed DPLL Search for MaxSAT (short paper) Lukas Kroc, Ashish Sabharwal, Bart Selman Cornell University SAT-09 Conference Swansea, U.K. July 3, 2009.
Themes of Presentations Rule-based systems/expert systems (Catie) Software Engineering (Khansiri) Fuzzy Logic (Mark) Configuration Systems (Sudhan) *
OPTIMIZATION WITH PARITY CONSTRAINTS: FROM BINARY CODES TO DISCRETE INTEGRATION Stefano Ermon*, Carla P. Gomes*, Ashish Sabharwal +, and Bart Selman* *Cornell.
1 MCMC Style Sampling / Counting for SAT Can we extend SAT/CSP techniques to solve harder counting/sampling problems? Such an extension would lead us to.
Performing Bayesian Inference by Weighted Model Counting Tian Sang, Paul Beame, and Henry Kautz Department of Computer Science & Engineering University.
INTRODUCTION TO ARTIFICIAL INTELLIGENCE COS302 MICHAEL L. LITTMAN FALL 2001 Satisfiability.
Explorations in Artificial Intelligence Prof. Carla P. Gomes Module 3 Logic Representations (Part 2)
Solvers for the Problem of Boolean Satisfiability (SAT) Will Klieber Aug 31, 2011 TexPoint fonts used in EMF. Read the TexPoint manual before you.
Week 10Complexity of Algorithms1 Hard Computational Problems Some computational problems are hard Despite a numerous attempts we do not know any efficient.
Heavy-Tailed Phenomena in Satisfiability and Constraint Satisfaction Problems by Carla P. Gomes, Bart Selman, Nuno Crato and henry Kautz Presented by Yunho.
Survey Propagation. Outline Survey Propagation: an algorithm for satisfiability 1 – Warning Propagation – Belief Propagation – Survey Propagation Survey.
U NIFORM S OLUTION S AMPLING U SING A C ONSTRAINT S OLVER A S AN O RACLE Stefano Ermon Cornell University August 16, 2012 Joint work with Carla P. Gomes.
1 The Theory of NP-Completeness 2 Cook ’ s Theorem (1971) Prof. Cook Toronto U. Receiving Turing Award (1982) Discussing difficult problems: worst case.
Learning With Bayesian Networks Markus Kalisch ETH Zürich.
CSE 589 Part VI. Reading Skiena, Sections 5.5 and 6.8 CLR, chapter 37.
Explorations in Artificial Intelligence Prof. Carla P. Gomes Module Logic Representations.
NP-COMPLETE PROBLEMS. Admin  Two more assignments…  No office hours on tomorrow.
1 Mean Field and Variational Methods finishing off Graphical Models – Carlos Guestrin Carnegie Mellon University November 5 th, 2008 Readings: K&F:
SAT 2009 Ashish Sabharwal Backdoors in the Context of Learning (short paper) Bistra Dilkina, Carla P. Gomes, Ashish Sabharwal Cornell University SAT-09.
CPSC 422, Lecture 21Slide 1 Intelligent Systems (AI-2) Computer Science cpsc422, Lecture 21 Oct, 30, 2015 Slide credit: some slides adapted from Stuart.
Review of Propositional Logic Syntax
Balance and Filtering in Structured Satisfiability Problems Henry Kautz University of Washington joint work with Yongshao Ruan (UW), Dimitris Achlioptas.
Probabilistic and Logical Inference Methods for Model Counting and Sampling Bart Selman with Lukas Kroc, Ashish Sabharwal, and Carla P. Gomes Cornell University.
Satisfiability and SAT Solvers CS 270 Math Foundations of CS Jeremy Johnson.
A COURSE ON PROBABILISTIC DATABASES Dan Suciu University of Washington June, 2014Probabilistic Databases - Dan Suciu 1.
Dana Nau: Lecture slides for Automated Planning Licensed under the Creative Commons Attribution-NonCommercial-ShareAlike License:
1 P NP P^#P PSPACE NP-complete: SAT, propositional reasoning, scheduling, graph coloring, puzzles, … PSPACE-complete: QBF, planning, chess (bounded), …
Hardware Acceleration of A Boolean Satisfiability Solver
Inference and search for the propositional satisfiability problem
A New Algorithm for Computing Upper Bounds for Functional EmajSAT
Samuel Luckenbill1, Ju-Yueh Lee2, Yu Hu3, Rupak Majumdar1, and Lei He2
Queries with Difference on Probabilistic Databases
Emergence of Intelligent Machines: Challenges and Opportunities
Chapter 11 Limitations of Algorithm Power
Expectation-Maximization & Belief Propagation
Knowledge Compilation: Representations and Lower Bounds
Presentation transcript:

Solution Counting Methods for Combinatorial Problems Ashish Sabharwal [ Cornell University] Based on joint work with: Carla Gomes, Willem-Jan van Hoeve, Lukas Kroc, Bart Selman INFORMS, Oct 2008, Washington, D.C.

INFORMS-082 Context  Constraint Satisfaction Problems (CSPs)  In particular, Boolean Satisfiability or SAT :  Given a Boolean formula F in conjunctive normal form e.g. F = (a or b) and (  a or  c or d) and (b or c) determine whether F is satisfiable  NP-complete  widely used in practice, e.g. in hardware & software verification, design automation, AI planning, … How many satisfying assignments does F have?  #F, the “model count” of F, the solution count of F  #SAT is #P-complete

INFORMS-083 Model Counting for SAT  Inspired by the success of SAT solvers, a lot of activity in the last few years in attacking the solution counting problem  Aside: “success of SAT” = scalability, industrial applications, black-box nature and standardized input making it ‘easy’ for users  Many different approaches, many different counting goals  A “zoo” of techniques!  This talk: to give a brief overview of these techniques, many of which are contributed by our group at Cornell  Further reading and refs: Model Counting chapter in the upcoming Handbook of Satisfiability (draft available on my webpage) – with Carla Gomes and Bart Selman

INFORMS-084 What shall we count? 0#F2N2N E.g., F has N=1000 variables and ≈ solutions 0#F Exact count Estimate, no guarantees Upper bound (appears hard!) Lower bound Strict “( ,  )” guarantee

INFORMS-085 Problem Space: why are upper bounds hard?  Number of solutions is often a miniscule fraction of the search space size  Limits our ability to reason about upper bounds  E.g., after having searched half the space, could still have potential solutions remaining in the worst case! (off by a factor of )  Probabilistic methods work better for lower bounds  E.g., if expected value = true count, Markov’s ineq. says, can’t get high numbers too often because 0’s can’t compensate enough  reverse Markov’s ineq. doesn’t help: can get low numbers too often because a single 2 N can compensate for a lot of low numbers! 0#F2N2N E.g., F has N=1000 variables and ≈ solutions

INFORMS-086 The “Zoo” of Counting Methods Exact methods Practical bounds with a guarantee Approximate methods Estimation without any guarantee Solution counting “Only” the count Count + many by-products DPLL-style backtrack search Knowledge compilation Using backtr. -free space Sampling + multipliers Sampling + randomization FPRAS: MCMC sampling FPT: branch-width, tree-width,… XOR streamlining (randomized) Backtr. search + randomization + statistics Belief prop. + randomization Note: not an exhaustive listing L LL L U U

I. Exact Methods Exact methods “Only” the count Count + many by-products DPLL-style backtrack search Knowledge compilation FPT: branch-width, tree-width,… [“CDP”, Birnbaul-Lozinskii-99] [“relsat”, Bayardo-Pehoushek-00] [“cachet”, Sang et al-04] [“sharpSAT”, Thurley-06] [tree-width: Gottlob-Scarcello-Sideri-02] [branch-width: Bacchus-Dalmao-Pitassi-03] [cluster-width: Fischer-Makowsky-Ravve-08]

INFORMS-088 Knowledge Compilation for Counting  Main idea: convert F into a different “form” from which one can easily read off the solution count (and many other quantities of interest)  d-DNNF: Deterministic, Decomposable Negation Normal Form  Think of the formula as a directed acyclic graph (DAG)  Negations allowed only at the leaves (NNF)  Children of AND node don’t share any variables (different “components”)  Children of OR node don’t share any solutions  Once converted to d-DNNF, can answer many queries in linear time  Satisfiability, tautology, logical equivalence, solution counts, …  Any query that a BDD could answer  Our recent result: can count number of “clusters” of solutions – how many different kinds/families of solutions are there? [DNNF, “c2d”, Darwiche et al ] can multiply the counts can add the counts [To appear in NIPS-08]

II. Approximate Methods Practical bounds with a guarantee Approximate methods Estimation without any guarantee Using backtr. -free space Sampling + multipliers Sampling + randomization XOR streamlining (randomized) Backtr. search + randomization + statistics Belief prop. + randomization LL L U L FPRAS: MCMC sampling U [Karp-Luby-85] [Karp-Luby-Madras89] [“SampleMinisat”, Gogate-Dechter-07] [“MiniCount”, CPAIOR-08]

INFORMS-0810 XOR Streamlining for Bounds on #F  Main idea: rather than modifying the algorithm for solving, modify the problem, run the solver, deduce the count  Randomized algorithm, expected value = true count  Can be converted into bounds with correctness guarantees  Lower bounds easier in practice (XORs of any “length” work)  Upper bounds possible but not so easy  Empirical evidence: can get by with “very short” XORs  Can be extended to general CSPs Streamlined formula CNF formula Random XOR constraints Off-the-shelf SAT Solver Model count [“Mbound”, AAAI-06] [SAT-07] [AAAI-07; see Willem’s talk] ideal when systematic search works well!

INFORMS-0811 Sampling for Estimates + Lower Bound  Main idea: “find” a balanced variable – one that appears roughly equally as True and as False in solutions; fix to one value, count that sub-problem, re-scale with appropriate multiplier  Finding balanced variables not so easy  Use solution sampling: ideal when local search works well!  Use Belief Propagation for “marginal” prob. estimates: ideal when message passing works well!  Randomize the process: expected value = true count, as before!  Great lower bounds, but variance too high for good upper bounds x=? TF 40% of solutions 60% of solutions E.g., count #F| x=T, scale up by factor 100/60 [“ApproxCount”, Wei-Selman-05] [“BPCount”, CPAIOR-08] [“SampleCount”, IJCAI-07]

INFORMS-0812 The “Zoo” of Counting Methods Exact methods Practical bounds with a guarantee Approximate methods Estimation without any guarantee Solution counting “Only” the count Count + many by-products DPLL-style backtrack search Knowledge compilation Using backtr. -free space Sampling + multipliers Sampling + randomization FPRAS: MCMC sampling FPT: branch-width, tree-width,… XOR streamlining (randomized) Backtr. search + randomization + statistics Belief prop. + randomization Note: not an exhaustive listing L LL L U U