Lappoon R. Tang

Assistant Professor of Computer Sciences, University of Texas at Brownsville

Ph.D. in Computer Sciences, University of Texas at Austin, 2003
M.S. in Computer Sciences, University of Texas at Austin, 1998
B.S. in Computer Sciences, University of Texas at Austin, 1995

Email: my_first_name.my_last_name@utb.edu 8-)
Phone: 1-956-882-6771

Research Interests

Professor Tang's area of research is in Machine Learning. It concerns building systems that can improve their performance with experience. He is particularly interested in learning in logic (e.g. first order logic). Learning in such a framework allows the use of background knowledge and learning is incremental. However, learning in full fledged first-order logic (FOL) is generally very inefficient. Classification in first order logic is also undecidable in general since FOL is undeciable (despite that it is arguable one can limit his learning to a restricted subset of FOL that is decidable). On the other hand, Horn-clause first order logic offers a much more attractive and viable alternative -- it has a non-trivial amount of expressive power but yet learning is generally more feasible (PAC-learning wise). His interest in learning in this framework has led him to contributing (however little) to a still growing subfield in Machine Learning called Inductive Logic Programming (ILP).

His previous works in ILP include its application to Natural Language Understanding for automated construction of natural language interfaces (NLI) for databases, and Link Discovery for building counter-terrorism applications (a huge data mining problem). He is currently working on building reliable NLI (i.e. one that is guaranteed to deliver only correct answers to the user) using a hybrid approach that combines the strength of machine learning and non-machine learning approaches. He has recently become a member of the LIGO Scientific Collaboration and is involved in data mining of astronomical data. His involvement in the project led him to developing new clustering algorithms. His current focus is in tackling the problem of cluster validation using a homogeneity detection approach (i.e. one that detects if a cluster is homogeneous) instead of traditional approaches that employ a cluster validation index where the problem is then basically treated as an optimization problem.

His other areas of interest also include Genetic Programming, and Functional Genomics in Bioinformatics.

Here is his research statement (somewhat outdated at the moment but it will be updated soon, sorry for any inconvenience). :-) .

Courses

COSC 4355 and COSC 5355: Expert Systems