Question about delta

[tutuaixizao]

The pseudocode in the paper describes that under a certain delta value, when the output results of the classifier are all zero, the delta value will be reduced by half, but I don’t seem to see this step in the code. Could you please explain this? I still occasionally encounter this problem when using your code to run experiments under different gamma values or different networks.

[Jianbo-Lab]

It is implemented in the geometric progression algorithm. The default uses binary search to find size. you may try geometric progression, and see how it performs for your networks.