Achievement

The formal core of the training program is a set of Formal Methods courses on mathematical and computational methods in cognitive science. A new such course on structured statistical models of inference was developed. Students learned fundamental concepts and techniques of algorithmic learning, including maximum likelihood and Bayesian estimation; marginalization over hidden structure; iterative approximation by expectation-maximization (EM); and maximum-margin separation. Specific topics covered by the course included the set-theoretic foundations of probability theory; constructing and training probabilistic generative models of natural language (sound systems, words, syntactic grammars) and other cognitive domains (knowledge of natural kinds, taxonomic knowledge, color categories); and application of support vector machines (SVM) and other types of classifiers to multi-voxel pattern analysis of functional Magnetic Resonance Imaging data.