Leveraging Highly Approximated Multipliers in DNN Inference

TL;DR

This paper introduces a control variate approximation method that allows the use of highly approximate multipliers in DNN inference, significantly reducing power consumption while maintaining accuracy without retraining.

Contribution

The authors propose a novel control variate technique that enables effective use of approximate multipliers in DNNs without retraining, improving accuracy and power efficiency.

Findings

Achieves 45% power reduction with less than 1% accuracy loss.

Improves accuracy by 1.9x over approximate designs without the technique.

Demonstrates effectiveness across six different DNNs and multiple multipliers.

Abstract

In this work, we present a control variate approximation technique that enables the exploitation of highly approximate multipliers in Deep Neural Network (DNN) accelerators. Our approach does not require retraining and significantly decreases the induced error due to approximate multiplications, improving the overall inference accuracy. As a result, our approach enables satisfying tight accuracy loss constraints while boosting the power savings. Our experimental evaluation, across six different DNNs and several approximate multipliers, demonstrates the versatility of our approach and shows that compared to the accurate design, our control variate approximation achieves the same performance, 45% power reduction, and less than 1% average accuracy loss. Compared to the corresponding approximate designs without using our technique, our approach improves the accuracy by 1.9x on average.