MinConvNets: A new class of multiplication-less Neural Networks

TL;DR

MinConvNets introduce a novel neural network architecture that replaces multiplications with simpler minimum comparator operations, significantly reducing computational complexity for deployment on embedded devices.

Contribution

The paper proposes a new class of neural networks that approximate multiplications with minimum operations, along with a methodology and training method to maintain accuracy.

Findings

MinConvNets can replace multipliers with minimum operations under certain statistical constraints.

Transfer learning enables MinConvNets to achieve comparable precision to traditional CNNs.

Hardware implementation of MinConvNets is simpler and more efficient.

Abstract

Convolutional Neural Networks have achieved unprecedented success in image classification, recognition, or detection applications. However, their large-scale deployment in embedded devices is still limited by the huge computational requirements, i.e., millions of MAC operations per layer. In this article, MinConvNets where the multiplications in the forward propagation are approximated by minimum comparator operations are introduced. Hardware implementation of minimum operation is much simpler than multipliers. Firstly, a methodology to find approximate operations based on statistical correlation is presented. We show that it is possible to replace multipliers by minimum operations in the forward propagation under certain constraints, i.e. given similar mean and variances of the feature and the weight vectors. A modified training method which guarantees the above constraints is proposed. And it is shown that equivalent precision can be achieved during inference with MinConvNets by using transfer learning from well trained exact CNNs.