Incremental Network Quantization: Towards Lossless CNNs with\n Low-Precision Weights

This paper presents incremental network quantization (INQ), a novel method,\ntargeting to efficiently convert any pre-trained full-precision convolutional\nneural network (CNN) model into a low-precision version whose weights are\nconstrained to be either powers of two or zero. Unlike existing methods which\nare struggled in noticeable accuracy loss, our INQ has the potential to resolve\nthis issue, as benefiting from two innovations. On one hand, we introduce three\ninterdependent operations, namely weight partition, group-wise quantization and\nre-training. A well-proven measure is employed to divide the weights in each\nlayer of a pre-trained CNN model into two disjoint groups. The weights in the\nfirst…