Efficient Exact Gradient Update for training Deep Networks with Very Large Sparse Targets
Mila > Publication > Efficient Implementations of Deep Nets