3 ms·Discrete Variational Autoencoders24 points by rch 10y agojostmey 10y agoTrying to wrap my mind around the paper. Could be a really important finding. Should now allow for efficient use of discrete variables in deep neural network models.