More details on SparseCtrl's training

[asheroin]

Thanks for this newly imprestive works but there still remains some confusions for me while training my own SparseCtrl networks. For example, the traning iterations, the beta schedule, the sample_size, the traning tricks and so on.

I had noticed that more details could be found in the supplymentary meterial. Howerver, it seems to have not been released currently.

Hoping for more details to be added to the README files or the release of the supplymentary meterial!