CS 289, Fall 2004, Stuart Russell
Reading List for Knowledge Representation and Reasoning

Book

Week 1 (8/30): AI, knowledge-based systems, propositional logic

• AIMA2e Ch.2 (skim)
  A general introduction to intelligent agents, the kinds of environments they work in, and the role of knowledge in agent design
• McCarthy, J., ``Programs with common sense.'' In Proceedings of the Symposium on Mechanisation of Thought Processes, Her Majesty's Stationery Office, 1958. [pdf]
  Probably the eaeliest paper advocating the declarative approach to building intelligent systems. Includes contemporary commentary and footnotes added more recently by McCarthy.
• AIMA2e Ch.7.1-7.5
  Knowledge-based agents, logic in general, and propositional logic in particular. Inference algorithms for propositional theorem-proving (resolution, forward and backward chaining).

Week 2 (9/6): Propositional logic contd.: Efficient inference

Required reading

• AIMA2e Ch.7.6.
  Model-checking algorithms for Boolean satisfability (DPLL and WalkSAT).
• Joao P. Marques-Silva, ``An Overview of Backtrack Search Satisfiability Algorithms.'' In Fifth International Symposium on Artificial Intelligence and Mathematics, 1998. [pdf]
  Summarizes a number of improvements to the basic DPLL algorithm, mostly as implemented in the author's GRASP solver.
• L. Zhang and S. Malik, ``The Quest for Efficient Boolean Satisfiability Solvers.'' In Proceedings of 8th International Conference on Computer Aided Deduction, 2002. [pdf]
  Describes most of the ideas underlying the CHAFF family of solvers, currently the fastest general purpose Boolean satisfiability solvers.
• U. Schoning, ``A probabilistic algorithm for k-SAT and constraint satisfaction problems.'' In FOCS-99. [pdf]
  Describes a simple stochastic satisfiability algorithm that is easy to analyze and runs in (4/3)^n time; the algorithm uses repeated solution attempts from random initial states and ends each attempt after 3n steps.

• Stephen A. Cook and David G. Mitchell, ``Finding Hard Instances of the Satsifiability Problem: A Survey.'' In Du, D., Gu, J., and Pardalos, P. (Eds.), Satisfiability Problem: Theory and Applications. American Mathematical Society, 1997. [pdf]
  Analysis of progress in propositional satisfiability algorithms and in our understanding of the hardness of randomly generated problems, including discussion of the phase transition phenomenon.
• Evgeny Dantsin, Andreas Goerdt, Edward A. Hirsch, Ravi Kannan, Jon Kleinberg, Christos Papadimitriou, Prabhakar Raghavan, Uwe Schoning, ``A deterministic algorithm for satisfiability based on local search.'' In Theoretical Computer Science, 289, 2002. [pdf]
  A deterministic derivative of Schoning's algorithm; does a hillclimbing search within each of a set of balls that cover the entire space.
• R. Bayardo and R. Schrag, ``Using CSP Look-Back Techniques to Solve Real-World SAT Instances.'' In AAAI-97. [pdf]
  Methods for improving on DPLL-style backtracking search for SAT by caching (a.k.a. "learning") partial results.
• "M. Luby, A. Sinclair, and D. Zuckerman, ``Optimal speedup of Las Vegas algorithms.'' In Information Processing Letters, 47, 173-180, 1993. [pdf]
  Analysis of optimal restart strategies for randomized algorithms. Includes a universal strategy that is close to optimal for any unknown runtime distribution.
• Selman, B., Kautz, H., and McAllester, D., ``Ten Challenges in Propositional Reasoning and Search.'' In Proc. IJCAI-97. [pdf]
  Research challenges, both practical and theoretical, arising from the apparent success of simple stochastic algorithms for SAT.

Week 3 (9/13): QBFs, propositional agents.

• Marco Cadoli, Andrea Giovanardi, and Marco Schaerf, ``An algorithm to evaluate quantified Boolean formulae.'' In Proc. AAAI-98, 1998. [pdf]
  Describes QBFs and a simple extension of the DPLL algorithm for evaluating them.
• AIMA2e, Ch. 7.7
  Propositional agents based on theorem-proving and sequential Boolean circuits.
• Stanley J. Rosenschein and Leslie Pack Kaelbling, ``A Situated View of Representation and Control.'' Artificial Intelligence, 73(1-2), 149-173, 1995. [pdf]
  A broad introduction to a research agenda aimed at a science of situated agent design based on compiling declarative environment descriptions into sequential circuits that provably meet specifications for information-tracking and goal achievement.

• J. Rintanen, ``Improvements to the evaluation of quantified Boolean formulae.'' In Proc. IJCAI-99, 1999. [pdf]
  A method that improves on Cadoli's algorithm by partially instantiating the universally quantified variables in the QBF.

Week 5 (9/27): First-order logic: representation

• AIMA2e, Ch. 8
  Syntax and semantics of first-order logic, along with many simple examples of its use.
• AIMA2e, Ch. 9.1-9.2
  Inference by propositionalization; unification and lifting.

Week 6 (10/4): First-order logic: inference

• AIMA2e, Ch. 9.3-9.6
  Forward and backward chaining with definite clauses, and logic programming; theorem-proving by resolution, including completeness.
• Stickel, M. E., ``A Prolog technology theorem prover: a new exposition and implementation in Prolog.'' Technical Note 464, Artificial Intelligence Center, SRI International, Menlo Park, California, June 1989. [pdf]
  Explains how to modify Prolog so that it is sound and complete, using model-elimination theorem proving; retains the advantages of compilation and efficient implementation. This is the tech report version that includes full code listings; pages 1-20 are the paper itself.

• Sagonas, K., Swift, T., and Warren, D.S., ``XSB as an efficient deductive database engine.'' SIGMOD Record, 23(2), 442-453, 1994. [pdf]
  Description of an implenentation of tabled logic programming, which is polytime for Datalog, and a comparison to other deductive database systems.

Week 7 (10/11): Reasoning about action

• AIMA2e, Ch. 10.3
  Situation calculus and related formalisms as basic tools for reasoning about actions in first-order logic.
• AIMA2e, Ch. 11.1, 11.2, 11.5
  Planning algorithms; planning by reduction to propositional model-finding.
• Kautz, H., McAllester, D., and Selman, B., ``Encoding plans in propositional logic.'' In Proc. Fifth Int'l Conf. on Principles of Knowledge Representation and Reasoning (KR '96), Boston, MA, Morgan Kaufmann, 1996. [pdf]
  Describes how to represent planning problems in propositional logic, such that every model corresponds to a valid plan and vice versa. An excellent example of applied KR and the KR debugging cycle.

• Thielscher, M., ``Towards State Update Axioms: Reifying Successor State Axioms.'' In Proc. Logics in Artificial Intelligence. European Workshop, JELIA '98, Dagstuhl, Germany, Springer Verlag, 1998. [pdf]
  Description of solutions to the representational and inferential frame problems, including the fluent calculus.
• M. Ernst, T. Millstein, and D. Weld, ``Automatic SAT-Compilation of Planning Problems.'' In Proc. IJCAI-97, Nagoya, Japan, 1997. [pdf]
  Supplements the analysis of Kautz and Selman with a fully automatic SAT compiler for planning problems and empirical analysis of the solution cost with various encodings.

Week 8 (10/18): Reasoning about action contd.; general ontology.

• AIMA2e, Ch. 12.3, 12.4
  Basic descriptions of logical decision problems under nondeterminism and partial observability, and a discussion of state-space conditional planning.
• P. Bertoli, A. Cimatti, M. Roveri, P. Traverso, ``Planning in Nondeterministic Domains Under Partial Observability via Symbolic Model Checking.'' In Proc. IJCAI-01, Seattle, 2001.[pdf]
  Fast conditional planning using OBDDs to represent belief states (for OBDDs see supplemental readings).
• AIMA2e 10.1, 10.2, 10.3 (second half).
  The organization of a globally consistent ontology for general knowledge representation.

• R. E. Bryant, ``Graph-Based Algorithms for Boolean Function Manipulation.'' IEEE Transactions on Computers, C-35(8) 677-691, 1986.[pdf]
  The basic paper on Ordered Binary Decision Diagrams (OBDDs), a canonical form for propositional logic; not very readable.
• R. E. Bryant, ``Symbolic Boolean Manipulation with Ordered Binary Decision Diagrams.'' ACM Computing Surveys, 24(3), 293-318, 1992.[pdf]
  A somewhat more up-to-date and readable paper on OBDDs; material relating to circuit design can be skipped.
• Henrik Reif Andersen, ``An Introduction to Binary Decision Diagrams." Lecture notes, 1997.[pdf]
  A very thorough set of lecture notes on OBDDs, including a proof of canonicality.

Week 9 (10/25): Reasoning about knowledge and communication.

• AIMA2e 10.4.
  A simple approach to reasoning about knowledge based on sentences as strings.
• E. Davis, ``A First-Order Theory of Communicating First-Order Formulas.'' In Proc. KR-04, Whistler, B.C., 2001.[pdf]
  A provably consistent treatment of knowledge and communication based on situation calculus and a first-order embedding of modal logic.

• Joseph Y. Halpern, ``Reasoning about knowledge: a survey.'' In Handbook of Logic in Artificial Intelligence and Logic Programming, Vol. 4, D. Gabbay, C. J. Hogger, and J. A. Robinson, eds., Oxford University Press, 1995.[pdf]
  A fairly thorough survey of methods for reasoning about knowledge, including a good treatment of modal logic, on which Davis's paper is based.

Week 10 (11/1): Bayesian networks: representation and inference.

• AIMA2e Ch. 13 (review) and 14 (except 14.6).
  Probability theory (review). Syntax and semantics of Bayes nets; exact and approximate inference. We will focus mainly on approximate inference, particularly MCMC.

• F. Bacchus, S. Dalmao, and T. Pitassi, ``Value : Bayesian inference via backtracking search.'' In Proc. UAI-03, 2003. [pdf]
  Develops an alternative approach to Bayes net inference based on DPLL-style backtracking and caching.
• A. Darwiche, ``Recursive conditioning.'' Artificial Intelligence, 126, 2001. [pdf]
  An "anyspace" algorithm for Bayes net inference, based on branch decomposition of the network.

Week 11 (11/8): Bayesian networks contd.

Week 12 (11/15): Qualitative probabilistic networks.; temporal probability models.

• M. Wellman, ``Fundamental concepts of qualitative probabilistic networks.'' Artificial Intelligence, 44, 1990. [pdf]
  The essential introduction to the ideas and mathematics of qualitative monotonicity constraints on conditional distributions and their importance to representation and decision making.
• AIMA2e Ch. 15.1-5
  General introduction to the representations and inference tasks for temporal models. The central ideas are developed first independently of any particular model family, and then details of HMMs, Kalman filters, and dynamic bayes nets are introduced.

Week 13 (11/22): Temporal probability models contd: activity monitoring and robotics.

Required reading

• AIMA2e Ch. 25.3
  Application of filtering to robotic localization and mapping
• K. Murphy and S. Russell, ``Rao-Blackwellised Particle Filtering for Dynamic Bayesian Networks.'' In Sequential Monte Carlo Methods in Practice, A. Doucet, N. de Freitas and N.J. Gordon (eds), Springer-Verlag, 2001. [pdf]
  Describes a simple method for decomposing certain filtering problems, enabling much larger problems to be handled. Illustrated on the robotic "simultaneous localization and mapping" (SLAM) problem..
• Lin Liao, Dieter Fox, and Henry Kautz, ``Learning and Inferring Transportation Routines.'' In Proc. AAAI-04, 2004. [pdf]
  A promising application of DBNs and RBPF for understanding what people are doing and helping them if they get lost.

• M. Montemerlo, S. Thrun, D. Koller, and B. Wegbreit, ``FastSLAM: A factored solution to the simultaneous localization and mapping problem.'' In Proc. AAAI-02, 2002. [pdf]
  An RBPF algorithm for landmark-based SLAM, with impressive demonstration of efficacy on large maps.
• M. Montemerlo, S. Thrun, D. Koller, and B. Wegbreit. ``FastSLAM 2.0: An improved particle filtering algorithm for simultaneous localization and mapping that provably converges.'' In Proc. IJCAI-03, 2003. [pdf]
  Improves on FastSLAM by generating particles conditioned on the new observation; also proves convergence to the exact map in the limit.

Week 14 (11/29): First-order probabilistic languages

• AIMA2e Ch. 14.6
  Introduction to the basic ideas of FOPLs and RPMs in particular.
• Hanna Pasula, Bhaskara Marthi, Brian Milch, Stuart Russell, and Ilya Shpitser, ``Identity Uncertainty and Citation Matching.'' In Advances in Neural Information Processing Systems 15, 2003. [pdf]
  Extension of RPMs to include relational and identity uncertainty, with application to identifying the papers referred to by citations.

• L. Getoor, N. Friedman, D. Koller, and A. Pfeffer, ``Learning Probabilistic Relational Models.'' In S. Dzeroski and N. Lavrac, Eds., Relational Data Mining, Springer-Verlag, 2001. [pdf]
  Probably the most readable introduction to RPMs appears in Section 1.4 of this paper.
• N. Friedman, L. Getoor, D. Koller and A. Pfeffer, ``Learning Probabilistic Relational Models.'' In Proc. IJCAI-99, 1999. [pdf]
  Section 4.1 of this paper includes the general analysis of acyclicity conditions for RPMs, missing from the book chapter version above..
• A. Pfeffer and D. Koller, ``Semantics and Inference for Recursive Probability Models.'' In Proc. AAAI-00, 2000. [pdf]
  Gives a more complete treatment of the well-definedness of certain RPMs with infinite ancestor chains, relevant to the simple genetics example in AIMA2e 14.6.

Week 15 (12/6): First-order probabilistic languages contd.