Pulse Neuron Circuit Based on Tunneling Field Effect Transistor

Abstract

The invention provides a pulse neuron circuit based on a tunneling field effect transistor, and belongs to the technical field of pulse neuron in neuromorphic calculation. The circuit comprises a capacitor, a leakage and reset tube, a positive feedback tube, and two stages of inverters connected in series, wherein the capacitor is used for simulating the cytomembrane capacitance of a biological neuron and accumulating charges caused by the input post-synaptic current; the leakage and reset tube is an N type TFET device that provides a leakage and reset path for the charges accumulated on the capacitor; the positive feedback tube is a P type TFET device that supplements charges for the capacitor when the input of a first stage of inverter approaches to a logic threshold level thereof; and the two stages of inverters connected in series are composed of complementary TFETs for amplifying the voltage change at an input end. Compared with the implementation mode based on the traditional MOSFET, the pulse neuron circuit provided by the invention has lower hardware overhead, and can effectively reduce the short circuit energy consumption every time when the circuit releases the pulse, sothat the circuit has higher energy efficiency.

Description

technical field [0001] The invention relates to the physical realization of pulse neurons in neuromorphic computing, in particular to a pulse neuron circuit based on tunneling field effect transistors. Background technique [0002] Today's society is in the era of data explosion and artificial intelligence, and data processing and computing are facing huge challenges. The traditional Von Neumann computing architecture is limited by the slowdown of Moore's Law and the bottleneck of its own memory wall. The improvement of computing power tends to be saturated and the degree of intelligence is low. It is increasingly unable to meet the needs of massive data processing and intelligent tasks, and other solutions are needed. Program. Inspired by the brain, the neural network computing architecture came into being. In recent years, the artificial neural network represented by the deep neural network has achieved great success in the field of perceptual intelligence such as image ...

Images