/LSTM_CUDA

Implement LSTM purely with CUDA

Primary LanguageCuda

LSTM_CUDA

Implement LSTM purely with CUDA used to analyze sentiment

TODO

  1. Select Dataset for sentiment analysis
  2. Specify which packages we need
  3. Specify which parts of LSTM should be accelerated with CUDA
  4. Write python scipt to compare results

Datasets

https://www.kaggle.com/crowdflower/first-gop-debate-twitter-sentiment

Updates

  1. Finished GateLayer Forward Algorithm
  2. Know that four gates are objects of class GateLayer. The GateLayer and DenseLayer should be child of Basic Layer.
  3. Need to understand the step of backward update.