Performance Enhancement in Machine learning Algorithm using FinFET Technology

The present machine learning classifiers utilizes digital MOSFETS circuits, which consumes lot of power for its processing. To overcome the mentioned issue, Mixed-signal machine learning classification has been harnessed effectively. In this work, FinFET based classifier has been developed for the enhancement of machine learning performance in terms of area, power and accuracy. Here, SG (Short-Gate) FinFET is used to generate high-resolution multiplication operation which plays a critical role for predictors in the classifier circuit. A classifier functions based on the output of each predictor, which is built using 45 nm FinFET technology and runs at 550 MHz with a mild 1V power supply. The MNIST dataset is used to verify the proposed machine learning classifier. The suggested work was simulated using SPICE to examine its performance in terms of prediction accuracy, which was 90%, energy consumption, which was 25 pJ per classification, and area, which was 1,365 μm
²
.