Sparse Deep Neural Network training

We would like to experiments with optimization methods applied to DNN’s in order to obtain a sparse trained network,  i.e. a network with a small number of neurons or connections.