Introduction to Neural Networks

Abstract

Modelled on the cerebral cortex of the human brain, Artificial Neural Networks are presently being applied to solve elusive problems of AI viz, speech and image recognition, computer vision and adaptive control. The neural net topology, interconnections and ‘learning’ algorithms are topics of current research. This paper introduces this exciting and fast expanding field by reviewing some of the important neural networks of proven performance. A detailed explanation and simulation examples are presented for Hopfield net, Hamming net, Bidirectional Associative Memory, Temporal Associative Memory and Kohonen's network. Perceptions and multilayer perceptions are also explained in the light of their class discrimination capabilities. The classical pattern recognition algorithms vis-a-vis their neural network equivalents are also discussed.

Indexing Terms:

• Artificial neural networks
• Parallel distributed processing
• Hopfield net
• Hamming net
• Bidirectional associative memory
• Temporal associative memory
• Kohonen's network
• Perceptron
• multilayer perceptrons