amineremache/qbso-fs

Question of search area and reinforcement learning

kaintels opened this issue · 1 comments

Hello.
Thank you for sharing great work. I am conducting research using these algorithms.

I have several questions.

  1. I understood the search area (=region) is a set of solutions generated from the reference solution by flipping in a reference solution using n/flip. the size of the search area is equal to the current number of bees. (it increases based on the total number of bees).

for example, if n is 4 [0,0,0,0] and flip is 2, and reference solution is [0,0,0,1],
By n/flip, it moves symmetrically by 2 bits. so I understood the search area is [1,0,0,1].
Is it correct? I am not sure if I understood it properly.

  1. In addition, can this one applied for neural networks and replay memory such as DQN?

Best Regards.
Seungwoo

This problem has been solved. Thank you.