Professor Susan Epstein
Artificial intelligence (AI) develops programmed agents (systems) that match or outperform people’s abilities to make decisions, to learn, and to plan. To do so, AI develops algorithms and methodologies that sense a system’s environment, decide what to do given that data, and effect its chosen actions in its environment.
This is an introductory, graduate-level course on artificial intelligence. It emphasizes fast and clever search heuristics, thoughtful ways to represent knowledge, and incisive techniques that support rational decision making. Application areas will include game playing, natural language processing, and robotics.
List of topics
Introduction: foundation definitions, classic AI problems, and their solutions, knowledge representation
State-space search: uninformed search, heuristic (informed) search, local search
Constraint satisfaction: principles and practices
Machine learning: foundation definitions, computational learning theory, major paradigms
Planning: as search and as a reactive process
Inference: probabilistic and logical reasoning, empirical concerns and complexity
Introduction to more advanced topics (e.g., embodied cognition, cog- nitive architectures, autonomy)
Students are expected to have a solid background in the analysis of algorithms, proofs in propositional and first-order logic, discrete mathematics, and elementary probability.
Students who successfully complete this course will be able to:
Discuss the agent paradigm as the goal of an intelligent machine.
Describe state space search as a mechanism for problem solving, including optimal solutions and their complexity.
Explain the role of caching, reactivity, heuristics, and planning in state space search.
Define machine learning and describe the specifics of several prominent machine-learning methods (e.g., SVMs, decision trees, Bayes nets, artificial neural networks, genetic algorithms)
Evaluate the complexity of an approach to a specific problem and its realistic impact.
Describe and illustrate the role of constraint satisfaction in AI, with appropriate examples.
Discuss the role of probabilistic reasoning and mechanisms that employ it
Discuss the role of logical reasoning and mechanisms that support it
Grades will be based on:
Russell and Norvig, Artificial Intelligence: A Modern Approach, the third edition. Students will also be required to read a wide variety of assigned papers, and summarize and react to their content.
Long-term goal: empirical research
Empirical AI research addresses a real-world problem with appropriate knowledge representations and a reasoning methodology for it, identifies or constructs algorithms to address it, and implements, tests, and evaluates alternative solution(s) to it. This course is intended to provide a solid foundation for empirical AI research.
About the instructor
Dr. Epstein is based at Hunter College and The Graduate Center, where she works on representing and learning to solve hard problems. Her collaborators span many disciplines, including psychology, geography, robotics, and linguistics. She is known for her pioneering work in representation and learning for game playing, path finding, and constraint satisfaction, all based in the context of her cognitive architecture, FORR. She is a past chair of The Cognitive Science Society and a current officer of SIGAI, the ACM’s Special Interest Group on Artificial Intelligence.