Regularizing Activation Distribution for Training Binarized Deep Networks
arXiv:1904.02823
Abstract
Binarized Neural Networks (BNNs) can significantly reduce the inference latency and energy consumption in resource-constrained devices due to their pure-logical computation and fewer memory accesses. However, training BNNs is difficult since the activation flow encounters degeneration, saturation, and gradient mismatch problems. Prior work alleviates these issues by increasing activation bits and adding floating-point scaling factors, thereby sacrificing BNN's energy efficiency. In this paper, we propose to use distribution loss to explicitly regularize the activation flow, and develop a framework to systematically formulate the loss. Our experiments show that the distribution loss can consistently improve the accuracy of BNNs without losing their energy benefits. Moreover, equipped with the proposed regularization, BNN training is shown to be robust to the selection of hyper-parameters including optimizer and learning rate.
References in corpus (7)
- Deep Learning with Limited Numerical Precision
- The Loss Surfaces of Multilayer Networks
- Self-Normalizing Neural Networks
- Towards Accurate Binary Convolutional Neural Network
- Apprentice: Using Knowledge Distillation Techniques To Improve Low-Precision Network Accuracy
- Deep Learning with Low Precision by Half-wave Gaussian Quantization
- AdaScale: Towards Real-time Video Object Detection Using Adaptive Scaling