Open GaoxiangLi opened 12 months ago
As the command you provided "python3 calypso.py --target-degree 1 --challenge-num 50000 --cut-length 64 --proc 1 --population 500 -landscape-evolution -aeomebic-reproduction", is that means 50000 CRP are generated to train at the beginning?
Based on the source code, for every 10 generations, 25% of the golden_challenge_set will be replaced by the newly generated CRP. That means the total CRP used in this process is "challenge-num" + 25% * (num_generation / 10) when "Landscape evolution".
If the newly generated CRP during the "Landscape evolution" are counted in the total CRP, it is difficult to have an even number for the total CRP. Could you share the setting of the parameter "challenge-num"?
Thanks for pointing it out.
This issue is not existent in the code running the hardware data. For simulation code, I'll push a PR. Ideally, what we computed, was a static set of N samples, and then using a randomised subset. But the current version indeed isn't that.
I'm self-assigning the issue. Will provide a more comprehensive version for the simulation and revert on this ticket.
@GaoxiangLi Yes