Fetching the paper…

Binarized Neural Networks: Training Deep Neural Networks with Weights and Activations Constrained to +1 or -1 · Around