Forward and Backward Information Retention for Accurate Binary Neural Networks

• Published in: Proceedings (IEEE Computer Society Conference on Computer Vision and Pattern Recognition. Online) pp. 2247 - 2256
• Main Authors: Qin, Haotong, Gong, Ruihao, Liu, Xianglong, Shen, Mingzhu, Wei, Ziran, Yu, Fengwei, Song, Jingkuan
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.01.2020
• Subjects: Backpropagation; Biological neural networks; Information entropy; Propagation losses; Quantization (signal); Training

Weight and activation binarization is an effective approach to deep neural network compression and can accelerate the inference by leveraging bitwise operations. Although many binarization methods have improved the accuracy of the model by minimizing the quantization error in forward propagation, th...