#HaLSTM: A LSTM Implementation in Halide
##Summary
##Background
####LSTM Neural Network
Long short-term memory (LSTM) is a recurrent neural network (RNN) architecture that is well-suited to learn from experience to classify, process and predict time series, especially when there are very long time lags of unknown size between important events. One of LSTM's successful application is in NLP area, where it is capable of fine-tuning the model's ability to handle long-term dependency. Compared to other famous deep learning models, LSTM has more complex topology in each neuron. This makes it more difficult to implement LSTM and exploit potential optimizations.
####Halide
##Approach
####Model Architecture
####Implementation in Halide
####Optimization on Performance
##Experiment
##References
##Collaborative Work Division