Energy-Efficient ConvNets Through Approximate Computing

TL;DR

This paper introduces approximate computing techniques to significantly reduce the energy consumption of ConvNets, enabling their deployment on energy-constrained devices without substantial accuracy loss.

Contribution

It presents novel system- and circuit-level methods for energy-efficient ConvNet accelerators that maintain high accuracy.

Findings

Achieves up to 30x energy savings without accuracy loss

Attains over 100x energy reduction at 99% accuracy

Demonstrates feasibility for embedded and wearable systems

Abstract

Recently ConvNets or convolutional neural networks (CNN) have come up as state-of-the-art classification and detection algorithms, achieving near-human performance in visual detection. However, ConvNet algorithms are typically very computation and memory intensive. In order to be able to embed ConvNet-based classification into wearable platforms and embedded systems such as smartphones or ubiquitous electronics for the internet-of-things, their energy consumption should be reduced drastically. This paper proposes methods based on approximate computing to reduce energy consumption in state-of-the-art ConvNet accelerators. By combining techniques both at the system- and circuit level, we can gain energy in the systems arithmetic: up to 30x without losing classification accuracy and more than 100x at 99% classification accuracy, compared to the commonly used 16-bit fixed point number format.