scripts for the experiments (run networks, build command lines for networks run, gather oar results, draw graphes from raw results, nystrom vs deepstrom example)

change in the dataset preprocessing + the nystrom layer has now only one matrix W instead of D and V

removed diagonal implementation of fastfood (only hadamard product version remaining) + implementation of the stacked fastfood for the hadamard product version - need command line execution support and then it will be good

First commit: implementation with tensorflow of deepfriedconvnet with non-adaptative fastfood and no fht - only one stack of fastfood - mnist dataset