Image Processing and Neural Networks Lab

Reference Text: Neural Networks; A Comprehensive Foundation,  by Simon Haykin

Professor: M. T. MANRY

A Neural Network is a nonlinear signal processor which:

(1) Has good approximation properties,

(2) Has basis functions which adapt during a training process,

(3) Closely mimics the optimal processor, and

(4) Can have its computational load varied incrementally 

Neural networks are widely used in document readers (by the postal service), power load forecasting (by electric utilities), financial analyses (by mutual fund companies), remote sensing (by oil service companies), and target recognition (by aerospace and defense companies). Additional applications of neural nets include nonlinear channel equalization in communication systems, speech recognition, face recognition and other biometric applications, recognition of tumors in medical imaging, and machine vision. 

In this course, students will:

(1) Learn neural network architectures (MLP, RBF, SVM, SOM), their training algorithms,

their relationships to optimal processors, and methods for predicting their capabilities.

(2) Produce neural net training software that is competitive with that commercially available.

(3) Learn how to evaluate nonlinear signal processors (both neural and conventional). The

evaluation metrics include mean-square error, classification error, free parameters, pattern

storage, number of multiplie to process an input, and multiplies per training iteration.

(4) Learn how neural nets are applied in character recognition, power load forecasting,

financial analyses, medical imaging, and remote sensing.  

Partial lecture notes are available at http://www-ee.uta.edu/EEweb/ip/index.html. In successive computer projects, students will develop increasingly effective training algorithms for feed-forward neural networks. An extra credit project is to assemble the computer projects into a user-friendly software package. Prerequisite: EE5350 (DSP) or concurrent registration. 

Prerequisites: None.

Click here to email any questions or comments about this web site.

Copyright © 2004 Image Processing and Neural Networks Lab
Last modified: March 06, 2007