Efficient Exact Gradient Update for training Deep Networks with Very Large Sparse Targets
Posted on25/12/2015
An important class of problems involves training deep neural networks with sparse prediction targets of very high dimension D. These occur naturally... Read More