CS 594 - Neural Information Processing
Spring 1998 - Bruce MacLennan

If you want a graduate-level neural net course, you should take this one, since there may not be another graduate- or undergraduate-level course before Fall 1999 or Spring 2000, if then.

Contact Information

Instructor:
Bruce MacLennan
Phone: 974-5067
Office: 110-A Ayres
Hours: 2:10-3:25 TR or make an appointment
Email: MacLennan@cs.utk.edu

Teaching Assistant:
Yushuai Lu
Hours: TBA
Email: lu@cs.utk.edu

This page: http://www.cs.utk.edu/~mclennan/Classes/594-S98/

Description

This course will be a comprehensive introduction to neural networks from an engineering and mathematical perspective. It will cover most of the material essential to anyone intending to use neural networks in their research.

Tentative List of Topics

A tentative list of topics is: overview of neural network models, overview of learning, correlation matrix memories, perceptrons, LMS algorithms, multilayer networks, RBF networks, models based in statistical physics, self-organizing systems (Hebbian, competitive learning, information theoretic approaches), temporal processing & neurodynamics.

Tentative Schedule

With a few exceptions we will try to do a chapter (or about 40 pages) per week (the first half-week counts as week 1). You will be expected to do the chapter reading before class, so that class time can be devoted to supplementary explanation.
  1. overview of neural network models (ch 1)
  2. overview of learning I (ch 2)
  3. overview of learning II (ch 2 ctu'd)
    correlation matrix memories (ch 3)
  4. perceptrons (ch 4)
    LMS algorithms (ch 5)
  5. multilayer networks I (ch 6)
  6. multilayer networks II (ch 6 ctu'd)
  7. multilayer networks III (ch 6 ctu'd)
  8. RBF networks (ch 7)
  9. models based in statistical physics (ch 8)
  10. self-organizing systems I: Hebbian (ch 9)
  11. self-organizing systems II: competitive learning (ch 10)
  12. self-organizing systems III: information theoretic approaches (ch 11)
  13. temporal processing (ch 13)
  14. neurodynamics (ch 14)
  15. supplementary topics

Prerequisites

Basic calculus (including elementary differential equations), linear algebra, probability and statistics. No previous experience with neural networks is necessary.

Text

Simon Haykin: Neural Networks: A Comprehensive Foundation (Macmillan 1994).

Grading

There will be homework and projects.

Meeting Time and Place

Tuesdays & Thursdays, 12:40 - 1:55 in BU 655.
Return to MacLennan's home page

Send mail to Bruce MacLennan / MacLennan@cs.utk.edu
Last updated: Tue Apr 21 14:58:42 EDT 1998