- Visual Analysis for Recurrent Neural Networks

LSTMVis allows you to interactively analyze the hidden state vectors of a recurrent neural network model, by simply selecting and comparing regions of the input. To demonstrate the system, we have provided a set of real example models and datasets to play with, including several word and character language models for text, music and code, a sequence auto-encoder, a German <-> English neural translation system, and a sentence summarization system. We recommend that you begin with the parentheses dataset, which has a simple and regular structure to demonstrate the use of the tool.