BinaryConnect: Training Deep Neural Networks with binary weights during propagations

TL;DR

BinaryConnect is a method for training deep neural networks using binary weights during propagation, significantly reducing computational complexity and acting as a regularizer, achieving near state-of-the-art results on several datasets.

Contribution

The paper introduces BinaryConnect, a novel approach that trains DNNs with binary weights during propagation while maintaining high-precision weights for updates.

Findings

Achieves near state-of-the-art results on MNIST, CIFAR-10, SVHN.

Reduces computational complexity by replacing multipliers with simple accumulations.

Acts as a regularizer similar to dropout.

Abstract

Deep Neural Networks (DNN) have achieved state-of-the-art results in a wide range of tasks, with the best results obtained with large training sets and large models. In the past, GPUs enabled these breakthroughs because of their greater computational speed. In the future, faster computation at both training and test time is likely to be crucial for further progress and for consumer applications on low-power devices. As a result, there is much interest in research and development of dedicated hardware for Deep Learning (DL). Binary weights, i.e., weights which are constrained to only two possible values (e.g. -1 or 1), would bring great benefits to specialized DL hardware by replacing many multiply-accumulate operations by simple accumulations, as multipliers are the most space and power-hungry components of the digital implementation of neural networks. We introduce BinaryConnect, a method which consists in training a DNN with binary weights during the forward and backward propagations, while retaining precision of the stored weights in which gradients are accumulated. Like other dropout schemes, we show that BinaryConnect acts as regularizer and we obtain near state-of-the-art results with BinaryConnect on the permutation-invariant MNIST, CIFAR-10 and SVHN.