可视化和理解循环网络

Jun, 2015

Visualizing and Understanding Recurrent Networks

Andrej Karpathy, Justin Johnson, Fei-Fei Li

TL;DR使用字符级语言模型作为可解释的测试平台，本研究分析了LSTM的表示、预测和错误类型，并揭示了其提高性能的长程结构依赖性的来源。

Abstract

recurrent neural networks (RNNs), and specifically a variant with long short-term memory (LSTM), are enjoying renewed interest as a result of successful applications in a wide range of machine learning problems t