Zhiyuan Tang 2016-08-22

Last week:

1. basicly, exploring the bridges between LSTM and GRU (convert LSTM to GRU to find what's important as GRU performs better than LSTM on WSJ ASR);

2. several attempts for residual learning on these two gated recurrent networks;

This week:

1. continue to find the links between LSTM and GRU;

2. residual learning continues;