Most current artificial neural networks exist only within software simulators running on conventional computers. Simulators can provide great flexibility, but require immensely powerful and costly hardware for even very small networks.

An artificial neural network implemented as a custom integrated circuit could operate many thousands of times faster than any simulator as each neuron can operate simultaneously. A significant problem with implementing neural networks in hardware is that larger networks require a great deal of silicon area, making them too costly to design and produce.

In this book, I test the effectiveness of a number of algorithms that reducing the size of a trained neural network while maintaining accuracy.

Contact Author: