Binarized Neural Networks: Training Deep Neural Networks with Weights and Activations Constrained to +1 or -1

TLDR

We introduce a method to train Binarized Neural Networks (BNNs) with binary weights and activations at run‑time. The method trains BNNs by using binary weights and activations to compute gradients during training and validates performance through experiments on Torch7 and Theano frameworks. BNNs drastically reduce memory and arithmetic operations, achieve near state‑of‑the‑art accuracy on MNIST, CIFAR‑10, and SVHN, run seven times faster on a GPU with a custom kernel, and the code is publicly available.

Abstract

We introduce a method to train Binarized Neural Networks (BNNs) - neural networks with binary weights and activations at run-time. At training-time the binary weights and activations are used for computing the parameters gradients. During the forward pass, BNNs drastically reduce memory size and accesses, and replace most arithmetic operations with bit-wise operations, which is expected to substantially improve power-efficiency. To validate the effectiveness of BNNs we conduct two sets of experiments on the Torch7 and Theano frameworks. On both, BNNs achieved nearly state-of-the-art results over the MNIST, CIFAR-10 and SVHN datasets. Last but not least, we wrote a binary matrix multiplication GPU kernel with which it is possible to run our MNIST BNN 7 times faster than with an unoptimized GPU kernel, without suffering any loss in classification accuracy. The code for training and running our BNNs is available on-line.