【论文阅读笔记】Deep Neural Network Compression with Single and Multiple Level Quantization

全文概括   本文是《Quantized Convolution Neural Networks for Mobile Devices》和《Incremental Network Quantization：Towards Lossless CNN with Low-Precision Weights》的思想结合。参考了前者的分层量化和k-means聚类共享权值，参考了后者的INQ思想，即同一层分块

>>阅读原文<<