Adaptive Neuron Activation Function with FGMOS Based Operational Transconductance Amplifier

In this paper, an operational transconductane amplifier (OTA) with tunable transconductance and threshold level, based on multiple input floating gate MOS (MIFGMOS) transistor is presented. This OTA has much wider linear range compared to CMOS based OTA. This circuit is suitable for analog signal processing and neural network applications. It can be used as neuron activation function (NAF) with programmable slope and threshold. Depending on the applications, in neural networks, it may be required to change the threshold levels and slope of the activation function. In the proposed circuit, these parameters can be easily changed by changing voltage at the control gate of FGMOS transistors. TSPICE simulation results using 2mum N-well process are presented.