S2-BNN: Bridging the Gap Between Self-Supervised Real and 1-bit Neural Networks via Guided Distribution Calibration

Abstract

Previous studies dominantly target at self-supervised learning on real-valued networks and have achieved many promising results. However, on the more challenging binary neural networks (BNNs), this task has not yet been fully explored in the community. In this paper, we focus on this more difficult scenario: learning networks where both weights and activations are binary, meanwhile, without any human annotated labels. We observe that the commonly used contrastive objective is not satisfying on BNNs for competitive accuracy, since the backbone network contains relatively limited capacity and representation ability. Hence instead of directly applying existing self-supervised methods, which cause a severe decline in performance, we present a novel guided learning paradigm from realvalued to distill binary networks on the final prediction distribution, to minimize the loss and obtain desirable accuracy. Our proposed method can boost the simple contrastive learning baseline by an absolute gain of 5.5∼15% on BNNs. We further reveal that it is difficult for BNNs to recover the similar predictive distributions as real-valued models when training without labels. Thus, how to calibrate them is key to address the degradation in performance. Extensive experiments are conducted on the large-scale Ima-geNet and downstream datasets. Our method achieves substantial improvement over the simple contrastive learning baseline, and is even comparable to many mainstream supervised BNN methods. Code will be made available.

Block Walsh-Hadamard Transform Based Binary Layers in Deep Neural Networks

“…Another approach is the SqueezeNet [28], which is designed as a small network with 1 × 1 convolutional filters. Yet another approach is to use binary weights in neural networks [29][30][31][32][33][34][35][36][37][38][39][40][41][42]. Since the weights of the neurons are binary, they can can be used to slim and accelerate networks in specialized hardware including compute-in-memory systems [43].…”

Fast Walsh-Hadamard Transform and Smooth-Thresholding Based Binary Layers in Deep Neural Networks

“…In [27] the neural network is slimmed by removing some layers from a well-developed model. Binary neural networks [28,29,30,31,32,33,34,35,36,37] showed that binary weights can be used to slim and accelerate neural networks.…”