Example script
python sMNIST.py --n_hid 100 --dt 0.042 --gamma 2 --epsilon 20 --gamma_range 2 --epsilon_range 2 --inp_scaling 1 --rho 9 --ron --trials 1 --topology orthogonal --sparsity 0.8
Model assessment can be triggered via --use_test
flag and by setting --trials 5
(to compute mean and standard deviation).
Please, note that, due to its structure, the sparsity level of the lower
matrix is different from all the others.
To get a matrix with a sparsity of 0.5, 0.6, 0.7, 0.8, 0.9 (as reported in the paper), you need to set the --sparsity
flag to 0, 0.2, 0.4, 0.6, 0.8 respectively.