Release 1.0

The first fully-functional TensorLog release.

All back-propagation is done in-house, so you're limited to softmax normalization.

We also do all our own row broadcasting for sparse matrix math, as well as several other sparse functions.