Current research in the Artificial Intelligence laboratory of the CSE department is focused on problems in natural language processing, vision, reasoning, and especially learning. Many of the projects involve the use of the sub-symbolic processing inherent in genetic algorithms and artificial neural (connectionist) networks.

Vision
We have ongoing projects in image compression, nonlinear dimensionality reduction, nonlinear edge operators, texture processing, face recognition, and object recognition. These projects use either neural networks or analytic algorithms.

Natural language processing
Our current projects investigate adaptive information retrieval, formal properties of natural language, statistical processing of large text corpora, word sense disambiguation and learning from linguistic instruction.

Reasoning
Work on automated reasoning spans a wide range, from complexity-theoretic investigations of the expressiveness of nonmonotonic logics, to architectures for multimedia expert systems, to modeling human arithmetic problem-solving.

Learning
Learning is a central theme of most work in the Artificial Intelligence laboratory. Our efforts range from including a developmental stage in genetic algorithms and using genetic algorithms as selectors for neural network architectures, to neural network approaches to learning dynamical behaviors, to statistical methods of extracting syntactic and semantic regularities from text, to expectation-maximization approaches to finding patterns in DNA databases, to explanation-based learning.

Much of this research is linked to an understanding of human cognition. As a result the AI faculty maintain close ties with UCSD researchers with expertise in these areas, including those in the Departments of Cognitive Science, Psychology, Linguistics, and Neurosciences. Research contacts have also been established between AI faculty and researchers working in similar areas at the Salk Institute and the Institute for Neural Computation.