Artificial Intelligence (AI)
Artificial Intelligence (AI) has a long history in our department, and currently supports a very dynamic program of research and education. Our educational curriculum provides a broad range of courses including introductory AI, automated planning, cognitive modeling, commonsense reasoning, evolutionary computation, game theory, machine learning, multi-agent systems, natural language processing, and neural computation. The AI group has consistently ranked high in external national assessments: for example, in the US News ranking of best graduate schools, our AI program is ranked 9th among all universities and 6th among public universities.
Many of our former students have gone on to very high levels of achievement. Examples include Vipin Kumar (PhD 1982), Fellow of the AAAS, ACM, and IEEE; Qiang Yang (PhD 1989), Fellow of the IEEE; Naresh Gupta (PhD 1993), Senior Vice President at Adobe; Lee Spector (PhD 1992), Fellow of the ISGEC; Gary Flake (PhD 1993), Microsoft Distinguished Engineer and well known author of The Computational Beauty of Nature; Narendra Ahuja (PhD 1979), Fellow of the IEEE, AAAI , SPIE, and ACM; and Granger Sutton (PhD 1992), whose software was used in the first-ever assembly of the complete whole genome of a free-living organism, the Haemophilus influenzae genome.
Early prominent members of the AI faculty included Laveen Kanal (heuristic search), Jack Minker (logic and databases), Charles Rieger (common sense reasoning and language), and Azriel Rosenfeld (computer vision). Following are brief profiles of our current faculty:
Hal Daumé III's research bridges the gap between natural language processing and machine learning. He is primarily interested in problems with complex internal structure, and for which background knowledge (eg., from linguistics) is available. In practice, he works primarily in the areas of Bayesian learning (particularly non-parametric methods), structured prediction and domain adaptation (with a focus on problems in language and biology). He's worked on applications ranging from document summarization to machine translation, from coreference resolution to biological factor analysis.
Bonnie Dorr does research on computational linguistics and co-directs the Computational Linguistics and Information Processing (CLIP) Laboratory. She is known for her research on linguistically-informed, semantically-inspired statistical models that are cross-linguistically applicable, yet practical to train and use. Her research group's systems for summarization, machine translation, and translation metrics have led to a first place standings in several NIST evaluation forums, including the Document Understanding Conference, the Machine Translation Evaluation Forum, and the Metrics MATR competition. She is past-president of the Association for Computational Linguistics and recipient of the NSF Presidential Faculty Fellowship Award, the Maryland's Distinguished Young Scientist Award, the Alfred P. Sloan Research Award, and the NSF Young Investigator Award. She is the author of Machine Translation: A View from the Lexicon.
Lise Getoor does research in machine learning and reasoning under uncertainty, especially as applied to structured and semi-structured data. She is one of the founders of the statistical relational learning (SRL) research area and is well-known for her work in probabilistic relational models and link mining. Her group Linqs does research in machine learning applied to networks and graphs, with a recent emphasis on social networks and social media. In addition, they work on related topics such as data integration and visual analytics. She is recipient of an NSF Career Award in 2008 and several best-paper awards, and was a Microsoft New Faculty Award finalist. More information can be found at http://www.cs.umd.edu/~getoor.
Dana Nau does research on automated planning and game theory. He is known for the discovery of “pathological” games (in which, counter-intuitively, looking ahead leads to worse decision-making), and for influential work on the theory and applications of automated planning. The algorithms he and his students have developed for AI planning, manufacturing planning, zero-sum games, and non-zero-sum games have won many awards and championships. His SHOP and SHOP2 planning systems have been downloaded more than 13,000 times and have been used in thousands of projects worldwide. He has more than 300 published papers and several best-paper awards, and is co-author of Automated Planning: Theory and Practice, the standard textbook on its topic. He is an elected Fellow of the Association for the Advancement of Artificial Intelligence.
Don Perlis studies commonsense reasoning—the ability to muddle through without expertise, even in unfamiliar situations—and the related areas of cognitive modeling and philosophy of mind and language. One of his main ongoing projects is the use of time-situated metacognitive computation for flexible, domain-general, self-adjusting autonomous systems, centered on a "commonsense-core hypothesis": that human-level intelligence hinges on a particular processing technique employing a concise set of basic error-handling rules, and that this can be programmed into automated systems as well. As this work progresses, Perlis applies it to various test domains, currently focusing on human-computer natural-language dialog.
Jim Reggia does research in the area of nature-inspired computing, including neural networks, genetic algorithms/programming, and artificial life. His research group's current/recent work includes such topics as developing integrated neurocomputational systems as models of human language processing and short term memory, symbolically interpreting the representations learned by neural networks, combining cause-effect reasoning with evolutionary search as a creativity-support tool for design, the use of swarm intelligence for problem solving, optimization, and self-assembly, and the use of genetic programming to evolve modular neural networks and self-replicating "machines". He is an elected Fellow of the American College of Medical Informatics and a Senior Member of the International Neural Network Society.
V. S. Subrahmanian works at the intersection of AI and databases, identifying methods by which to perform intelligent reasoning on massive, diversely represented data. His work encompasses logic and probabilistic methods. His work includes algorithms to reason about graph and network data (such as RDF-based semantic web data, algorithms to learn probabilistic models of behaviors from historical information, algorithms to predict possible future scenarios from large histories, as well as methods to analyze opinion expressed in text and extract information scalably from massive text archives and frameworks to identify (probabilistically) activities occurring both in video and transaction data. He has applied these methods to a variety of problems including learning models of terror group behavior and using them to make forecasts. Prof. Subrahmanian is an elected Fellow of both the American Association for the Advancement of Science and the Association for the Advancement of Artificial Intelligence.
In addition to the above faculty, the AI group has close ties with our department’s computer vision and spatial reasoning group, where faculty work in research areas that overlap with AI (Yiannis Aloimonos, Larry Davis, David Jacobs, Hanan Samet). Also involved with our AI program are faculty from other campus units, including John Horty (Philosophy), Sarit Kraus (Institute for Advanced Computer Studies, and Bar Ilan University’s Computer Science Department), Ugur Kuter (Institute for Advanced Computer Studies), Satyandra Gupta (Mechanical Engineering), William Rand (Business School), Philip Resnik (Linguistics), and Amy Weinberg (Linguistics).
