hrishikeshshekhar/Vanilla-RNN

Building a Vanilla RNN for sentiment analysis

Requirements

• numpy
• matplotlib

Setup

Install python packages

    pip install -r requirements.txt

Download glove word2vecs

wget http://nlp.stanford.edu/data/glove.6B.zip

Download imbd reviews data

wget https://ai.stanford.edu/~amaas/data/sentiment/aclImdb_v1.tar.gz
tar -xf imbd_reviews.tar.gz
mv aclImdb imbd_data

TODO

• Create a requirements.txt (Add install instructions)
• Make RNN class a separate file
• Display model params
• Feed forwarding batches and train on mini batches
• Read glove WordToVec
• Save and load weights
• Interactive command line for testing
• Experiment with better initializers (Xavier Initializer)
• Experiment with different optimizers (SGD, Batch GD, Momentum, RMSProp)
• Extract, analyse and train on IMBD Reviews
• Conduct study for hyper parameter tuning
• Compare results with Tensorflow's RNN, LSTM module
• Check for vanishing and exploding gradients
• Add regularization (Dropout, L1, L2)
• Create API for sentiment classification
• GPU support