The research in our lab is in machine learning. A branch of the artificial intelligence area of computer science, machine learning is the theory and practice of finding computational solutions to problems by training the computer to learn to identify them. Our work spans both the theoretical and experimental ends of the field.
With an Information Science program PhD student, Chih-Chung Kao, we are developing a Maximum Entropy training method for Artificial Neural Network (ANN) machine learning algorithms. The experimental work in our laboratory is done by groups comprised of the PI and students at all levels, both graduate and undergraduate.
The first area of experimental research in our lab is in the classification and prediction of local structure information in globular proteins. In collaboration with Prof. Jacquelyn Fetrow of The Scripps Research Institute, we developed a classification of the local structure in proteins, called Structural Building Blocks (SBBs). SBBs use machine learning methods, specifically (ANNs) and clustering algorithms, to analyze the local structure in proteins and infer six broad categories for local structure. This classification is more general than traditional secondary structure. Much of our current research is exploring the implications of SBB structures for side chain placement, local structure prediction, and the identification of structural motifs in loop regions of proteins.
Since graduate school, I have been intrigued by human language, and how computers can be used to both understand and recreate these amazing abilities. Stemming from debates in the mid 1980's on the nature of the human syntactic abilty and the power and relevance of artificial neural network models, I have been using the latter to understand and recreate aspects of the former.
Last Revised 10/05/01