EE 455/555, CS 410/510, & SySc 575
G.G. Lendaris, 725-4988, lendaris@sysc.pdx.edu
Fall, 1999 M W, 6:40-8:30 pm, SA 54
Neural Networks I
Neural networks is a computational and engineering methodology based on emulating how nature has implemented biological brain (in particular, the brain’s massively parallel and learning aspects). As such, it holds promise for significant impact on how important classes of scientific and engineering problems are solved. The objective of the two-term sequence is to have the students obtain a working knowledge of this forefront technology.
Neural Networks I (Fall term):
Covers basic ideas of the neural network (NN) methodology, a computing paradigm whose design is based on models taken from neurobiology and on the notion of "learning." A variety of NN architectures and associated computational algorithms for accomplishing learning are studied. Experiments with various of the available architectures are performed via a (commercial) simulation package. Students do a project on the simulator, or do a special programming project. Prerequisites: senior standing in EE or CS, or grad standing.Texts
:1. Neural Networks - A comprehensive foundation, Haykin, Simon, Prentice Hall 1999.
2. Neural Computing (tutorial volume of manual for the NeuralWorks simulation package), NeuralWare, Inc., 1993.
[Neural Networks II (Winter term): Focus is on applications of the neural network methodology appearing in the literature. Sufficient theoretical material in selected area(s) is covered, and then neural network applications to those areas are explored. A specific topic to be covered is Fuzzy Set Theory. Applications will include control systems, pattern recognition and other timely topics. Students do an application project of their choice (using the NN simulator). Prerequisite: Neural Networks I.
Texts:
To be specified.
Note:
The $80 "lab fee" charged for this course entitles the student to a set of diskettes for the neural network simulation package NeuralWorks Professional II/Plus (list price $1995), via a site licence agreement with NeuralWare, Inc. The User’s Guide is additional charge (approx. $75-80) [includes text #2 above].Schedule for Fall, 1999
.Topics to be covered include:
Intro/Overview; single-layer feed-forward networks (Perceptrons, Adaline); multi-layer feed-forward networks (including well-known Backpropagation algorithm); feedback networks/ associative memories (Hopfield, BAM, BSB); unsupervised learning/self-organizing networks (Competitive, Counterprop, LVQ, SOFM, ART). General ideas applicable to all will be discussed throughout the term, e.g., notions of performance criteria, network capacity, ability to generalize well, etc.
In addition to the reading assignments, there will be assignments on the simulator to do experimentation with the types of neural network architectures being studied, with a major project given during the last 3-4 weeks of the term, based on data and a problem context that I will provide. If a CS student prefers to define and carry out an appropriate programming task, that is negotiable.
In-class midterm and final exams.