An intelligent agent needs to be able to solve problems in its world. The ability to create representations of the domain of interest and reason with these representations is a key to intelligence. In this course we explore a variety of representation formalisms and the associated algorithms for reasoning. We start with a simple language of propositions, and move on to first order logic, and then to representations for reasoning about action, change, situations, and about other agents in incomplete information situations. This course is a companion to the course “Artificial Intelligence: Search Methods for Problem Solving” that was offered recently and the lectures for which are available online.
INTENDED AUDIENCE : BE/ME/MS/MSc/PhD studentsPREREQUISITES : Some exposure to formal languages, logic and programmingINDUSTRY SUPPORT : Software companies dealing with knowledge and reasoning, including the semantic web and semantic search.
COURSE LAYOUT Week 1: Introduction, Propositional Logic, Syntax and Semantics
Week 2: Proof Systems, Natural Deduction, Tableau Method, Resolution Method
Week 3: First Order Logic (FOL), Syntax and Semantics, Unification, Forward Chaining
Week 4: The Rete Algorithm, Rete example, Programming Rule Based Systems
Week 5: Representation in FOL, Categories and Properties, Reification, Event Calculus
Week 6: Conceptual Dependency (CD) Theory, Understanding Natural Language
Week 7: Deductive Retrieval, Backward Chaining, Logic Programming with Prolog
Week 8: Resolution Refutation in FOL, FOL with Equality, Complexity of Theorem Proving
Week 9: Semantic Nets, Frames, Scripts, Goals and Plans
Week 10: Description Logic (DL), Structure Matching, Classification
Week 11: Extensions of DL, The ALC Language, Inheritance in Taxonomies
Week 12: Default Reasoning, Circumscription, The Event Calculus Revisited
Week 13: Default Logic, Autoepistemic Logic, Epistemic Logic, Multi Agent Scenarios