CSC. 599.44    NEURAL COMPUTING    Spring 2001

Professor: Octavio Betancourt    Office: NAC 8/216A

Phone:     212-650-6180

Texts:   "Neural Networks: A comprehensive Foundation", 2nd, Ed., Simon Haykin. Macmillan, New York. 1999.

            "Neural Networks for pattern Recognition", Christopher M. Bishop, Oxford University press, 1995.

            "Pattern Recognition", S. Theodoridis and K. Koutroumbas. Academic Press, 1999.

Grading: 2 Projects:

    1. Multi-layered, feed forward network training, parity function calculation (40%)

    2. Application (Optical Character recognition or other) (60%)

Software: We will use Matlab 6 with Neural Network Toolbox.

Material to be covered:

    I. Introduction

        1. What is a Neural Network

        2. Models of a Neuron

        3. Neural Networks viewed as directed graphs

        4. Feedback

        5. Network Architectures

        6. Knowledge Representation

        7. Visualization Processes in Neural Networks

        8. Artificial Intelligence and Neural Networks

    II. Learning Process

        1. Error correction learning

        2. Hebbian Learning

        3. Competitive Learning

        4. Supervised Learning

        5. Unsupervised Learning

    III. The Perceptron

        1. Basic Considerations

        2. Performance Measure

        3. Maximum Likelihood Gaussian Qualifier

    IV. Least Mean Square Algorithm

        1. Method of Steepest Descent

        2. Least Mean Square Algorithm

    V. Multilayer Perceptrons

        1. Derivation of the Back-Propagation Algorithm

        2. The XOR Problem

        3. Generalization

        4. Cross Validation

        5. Approximation of Functions

        6. Supervised Learning of Probability Functions by multilayered Perceptrons

        7. Optical Character Recognition

        8. Speech Recognition