Binary Ensemble Neural Network: More Bits per Network or More Networks per Bit?
This addresses the accuracy loss issue in BNNs for efficient deep learning applications, representing a novel approach rather than an incremental improvement.
The paper tackles the problem of severe accuracy degradation in binary neural networks (BNNs) by proposing the Binary Ensemble Neural Network (BENN), which uses ensemble methods to improve performance with limited efficiency cost, resulting in BENN surpassing the accuracy of full-precision networks with the same architecture.
Binary neural networks (BNN) have been studied extensively since they run dramatically faster at lower memory and power consumption than floating-point networks, thanks to the efficiency of bit operations. However, contemporary BNNs whose weights and activations are both single bits suffer from severe accuracy degradation. To understand why, we investigate the representation ability, speed and bias/variance of BNNs through extensive experiments. We conclude that the error of BNNs is predominantly caused by the intrinsic instability (training time) and non-robustness (train & test time). Inspired by this investigation, we propose the Binary Ensemble Neural Network (BENN) which leverages ensemble methods to improve the performance of BNNs with limited efficiency cost. While ensemble techniques have been broadly believed to be only marginally helpful for strong classifiers such as deep neural networks, our analyses and experiments show that they are naturally a perfect fit to boost BNNs. We find that our BENN, which is faster and much more robust than state-of-the-art binary networks, can even surpass the accuracy of the full-precision floating number network with the same architecture.