BriefGPT.xyz
Sep, 2016
语音识别中门控循环神经网络的记忆可视化
Memory Visualization for Gated Recurrent Neural Networks in Speech Recognition
HTML
PDF
Zhiyuan Tang, Ying Shi, Dong Wang, Yang Feng, Shiyue Zhang
TL;DR
本文使用可视化技术研究了LSTM和GRU在语音识别任务中的行为，并提出两种简单而有效的网络结构修改：LSTM中的懒惰单元更新和残差学习的快捷连接。两种修改都使得网络更加易于理解和强大。
Abstract
recurrent neural networks
(RNNs) have shown clear superiority in sequence modeling, particularly the ones with gated units, such as long short-term memory (
lstm
) and gated recurrent unit (
→