BinaryConnect: Training Deep Neural Networks with binary weights during propagations

TLDR

Deep neural networks have achieved state‑of‑the‑art performance using large datasets and models, a progress largely enabled by GPUs, and future advances will require faster, low‑power computation, motivating research into specialized hardware such as binary‑weight networks that replace costly multipliers with simple accumulations. We introduce BinaryConnect, a method that trains a DNN with binary weights during forward and backward propagations while keeping high‑precision stored weights for gradient accumulation. BinaryConnect trains networks by using binary weights during propagation steps while maintaining full‑precision weights for updates, thereby reducing multiply‑accumulate operations to simple additions. BinaryConnect acts as a regularizer and achieves near state‑of‑the‑art accuracy on permutation‑invariant MNIST, CIFAR‑10, and SVHN.

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.