IEEE Computer Society

Chapter Home
Meeting Schedule
Recent Meetings
Chapter Information
Mailing List


Atlanta Chapter


May 22, 2001
Meeting Abstract


Higher-Order Neural Networks for Signal Prediction


Real-time signal prediction would enable significant advances in voice detection and fraud detection. This presentation is dedicated to explore several investigations for signal prediction that have the potential to yield real-time prediction hardware. Neural networks are inherently parallel algorithms and, as such, can be made to evaluate very quickly once trained for a specific task. A nonadaptive predictor must be designed to recognize and predict a broad range of signal conditions. However, an adaptive predictor can be simpler and potentially less sensitive to parameter variations since it evolves as signal conditions change. Several neural network approaches will be presented in conjunction with the multiple computer languages that have been used to speed up the training. Each approach utilizes the feedforward sigmoidal neural network trained with a backpropagation algorithm. For comparison, a linear predictor is also trained on the same data. Generally, the higher-order neural networks exhibit better performance characteristics than the linear predictor. These results also show that higher-order neural networks can be trained to predict 2-D and potentially 3-D signals, a feature that makes the future implementation of real-time prediction hardware feasible.


Dr. Wei Zhao is a senior member of the technical staff in the Science and Technology section at BellSouth. Wei has lived in Atlanta since the beginning of 2001. Prior to relocating in Atlanta, he worked for AT&T and lived in Northern Virginia for more than three years. Before he went to Virginia, he lived in the capital of the Peach State for six years. During this time, he received his Masters and Ph.D. degree in Electrical and Computer Engineering from Georgia Institute of Technology. Wei is a graduate of Shanghai Jiaotong University in Electrical and Computer Engineering. He has published several excellent papers in a variety of peer reviewed computer journals. Wei's interests include AIN, wireless, neural networks, high-speed signal processing, and oil painting.

Meeting Location and Time

May 22, 2001 at 7:00 p.m.
Corporate Values Room
Bell South Tower in midtown Atlanta
(near Crawford Long Hospital, the North Avenue Marta station and the Renaissance Hotel)

Lowell E. Reed
(770) 798-4940