Pattern Recognition and Neural Networks

Topics of the course will initially survey pattern recognition systems and components; decision theories and classification: discriminant functions: classical supervised and unsupervised learning methods, such as backpropagation, radial basis functions: clustering; feature extraction and dimensional reduction; sequential and hierarchical classification; Kohonen networks; Boltzman machines, principal components, and examples of applications. Modern concepts in learning will be introduced: nonparametric learning, reinforcement learning, mixtures models, belief networks, minimum description length, maximum likelihood, entropy methods, independent component analysis.

012503