Part I – Adaptive Signal Processing:

Adaptive filters have found many uses in today’s technology.   They are generally constructed as transversal digital filters with weighting coefficients automatically adjusted to minimize mean square error.  Convergence and learning speed are predictable and well understood.

 

In signal processing, applications include noise cancellation, deconvolution, and adaptive beamforming.  In telecommunications, applications include channel equalization and echo canceling.  In control systems, applications include inverse control and cancellation of plant disturbance.  In acoustics, applications include active control of sound and vibration.  Many of these subjects will be discussed.

Part I I – Adaptive Neural Networks

Although early forms of neural networks have existed since the 1960’s, advances in hardware, software, and neural algorithms have now brought  the technology to the fore.  This tutorial will describe the basic neural element and how such an element can be replicated and incorporated into a neural network capable of performing complex information processing tasks.

 

Neural networks can be trained and can self-learn to recognize patterns, to recognize speech, to predict weather from pressure patterns, and to perform control functions of considerable complexity.  Recent developments have led to a neural control system that has learned by itself to steer a trailer truck while backing to a loading platform.  The neural controller has solved a highly nonlinear control problem.  A video will be shown to demonstrate the “truck backer-upper” in action.

 

It is expected that this new technology will in the future play an important role in the control of power systems, of industrial plants, of  robotic systems, and in many other practical applications.