An approach for transistor implementation of neural networks and its various prospects and limitations

In signal processing neural network is a very efficient tool for implementing both linear and non-linear transfer function. But the system may not be so economic. On the other hand, the semiconductor device used may be unnecessary hefty for a small commercial application. So it would be sensible to implement it with transistor. Parameters need to be calculated by the traditional learning method used and then when it is done the work is to formulate the transfer function and then implement the transfer function with a analog circuit containing transistor, resistors, capacitors, inductors, and sometimes op-amps etc. The work may become quite hectic incase of formulation of highly complex neural network system. In such case the process can be made easy by calculating the parameters using a program based system. In the demonstration the system used are elementary and the same procedure repeatedly used can lead to highly complex and several layered neural network system.