Skip to content

Latest commit

 

History

History
1 lines (1 loc) · 337 Bytes

README.md

File metadata and controls

1 lines (1 loc) · 337 Bytes

In this ipython/jupyter tutorial I use Keras/Theano and an LSTM to predict sentiment for the IMDB dataset. This method is than compared empirically to TF-IDF and logistic regression. Finally, I pull the learned embeddings out of the model and analyze their structure with cosine search and T-SNE to demostrate what the LSTM has learned.