CMSC 691B Topic Ideas

Active learning

Suggested by Prof. Marie desJardins
Description: Machine learning techniques for selecting the most useful instance to label next.
Relevant publications: Machine Learning Journal, Journal of Machine Learning Research, AAAI, IJCAI, Journal of AI Research
Papers:

Cohn, Atlas, Ladner, "Improving generalization with active learning," Machine Learning 15(2): 201-221, 1994.
Cohn, Ghahramani, and Jordan, "Active learning with statistical models," JAIR 1996.
Tong and Koller, "Active learning for structure in Bayesian networks", IJCAI-01.

Plan explanation and summarization

Suggested by Prof. Marie desJardins
Description: Methods for automatically producing concise, informative summaries of plans produced by AI planning systems (particularly hierarchical task network planners)
Relevant publications: AAAI, IJCAI, Artificial Intelligence Journal, Journal of AI Research, AIPS
Papers:

Clement and Durfee, "Theory for coordinating concurrent hierarchical planning agents using summary information," AAAI-99
Weld, "An introduction to least-commitment planning," AI Magazine 1994
Mulvehill, "Plan comparison and summarization," AIPS-98 - Autobrief, http://www-2.cs.cmu.edu/Groups/sage/autobrief.html

Distributed Information Retrieval

Suggested by Prof. Charles Nicholas
Description: Techniques for dividing documents, indices, or metadata among a set of computers linked by a network.
Publications: SIGIR conference, CIKM conference, and sometimes VLDB. Journals include Information Retrieval, Information Processing and Managemnent.
Papers:

J. Callan. Distributed information retrieval. In W.B. Croft, editor, Advances in information retrieval, chapter 5, pages 127-150. Kluwer Academic Publishers, 2000.
The Impact of Database Selection on Distributed Searching Allison L. Powell, James C. French, Jamie Callan SIGIR 2000
L. Gravano, and H. Garcia-Molina. Generalizing GlOSS to Vector-Space databases and Broker Hierarchies. VLDB Conf., 1995.

Ebiquity Topics

Suggested by Prof. Tim Finin
See http://ebiquity.umbc.edu/v2.1/research/area/id/future/ for a list of project ideas related to the eBiquity lab.

Spam Filtering

Suggested by Prof. Charles Nicholas
Description: Everybody knows what spam is - email that is unsolicited and unwanted
Publications: USENIX Internet Technology conference, IEEE Internet Computing magazine, many web sites.
Papers:

Lots of papers on information filtering. There are likely also many papers that deal with spam as such.

Evaluation of Clustering Algorithms

Suggested by Prof. Charles Nicholas
Description: Clustering algorithms are used to find group data objects into sets, for which one data object can serve as a representative. But how can one tell if one clustering algorithm produces "better" clusters than another?
Publications: SIGIR conference, CIKM conference, many if not all data mining conferences, statistical literature on cluster analysis
Papers:

Peter Willet's survey from Information Processing and Management 1986 is seminal but now somewhat dated.

Many papers talk about k-means and the numerous variations.