PROJECT TITLE :

STT-SNN: A Spin-Transfer-Torque Based Soft-Limiting Non-Linear Neuron for Low-Power Artificial Neural Networks

ABSTRACT:

Recent years have witnessed growing interest in the employment of artificial neural networks (ANNs) for vision, classification, and inference problems. An artificial neuron sums N weighted inputs and passes the result through a non-linear transfer function. Large-scale ANNs impose terribly high computing necessities for training and classification, resulting in nice interest in the utilization of post-CMOS devices to appreciate them in an energy efficient manner. In this paper, we tend to propose a spin-transfer-torque (STT) device based on domain wall motion (DWM) magnetic strip which will efficiently implement a soft-limiting non-linear neuron (SNN) operating at ultra-low supply voltage and current. In distinction to previous spin-based neurons that can solely realize arduous-limiting transfer functions, the proposed STT-SNN displays a continual resistance amendment with varying input current, and will so be utilized to implement a soft-limiting neuron transfer operate. Soft-limiting neurons are greatly most well-liked to onerous-limiting ones because of their much improved modeling capability, which ends up in higher network accuracy and lower network complexity. We also gift an ANN hardware style using the proposed STT-SNNs and memristor crossbar arrays (MCA) as synapses. The ultra-low voltage operation of the magneto metallic STT-SNN enables the programmable MCA-synapses, computing analog-domain weighted summation of input voltages, to also operate at ultra-low voltage. We have a tendency to modeled the STT-SNN using micro-magnetic simulation and evaluated them using an ANN for character recognition. Comparisons with analog and digital CMOS neurons show that STT-SNNs can achieve around 2 orders of magnitude lower energy consumption.