《Batch Normalization Accelerating Deep Network Training by Reducing Internal Covariate Shift》閱讀筆記與實現

此文章爲轉載，原文地址在此 　　今年過年以前，MSRA和Google相繼在ImagenNet圖像識別數據集上報告他們的效果超越了人類水平，下面將分兩期介紹二者的算法細節。git 　　此次先講Google的這篇《Batch Normalization Accelerating Deep Network Training by Reducing Internal Covariate Shift》，主要

>>阅读原文<<