Prototype neural network hardware with analog electronic circuit

The present work realizes a model for hardware of neural network with four synaptic input links and corresponding weights constructed with operational amplifier. It can generate, in terms of dc voltage transfer characteristic, sigmoid functions of varied steepness and outputs for random variations of inputs, weights and threshold. The limitation in threshold variation is also estimated. Neural network hardware realization like the present one is unsuitable for VLSI implementation, but useful to discrete-component systems and has flexibility in power supply and output levels.