Sublinear Deep Learning Engine for CNNs

A probabilistic approach to reducing training time for CNNs with tensor sketching and locality-sensitive hashing.