Binarized Neural Networks: Training Neural Networks with Weights and Activations Constrained to +1 or −1
Recent results show that deep neural networks achieve excellent performance even when, during training, weights are quantized and projected to a binary representation.
Republish the Results of the paper( https://papers.nips.cc/paper/6573-binarized-neural-networks ) in pytorch.
Binarizing weights and activations for the RNNs works fine with the smaller datasets like training a set of words of diffrent languages and predicting to which country it belongs!. Binarizing weights and activations works fine with the smaller datasets and the results are pending for the PTB Dataset which is used in Natural Language Processing.