search

Matteo-Pietro-Pillitteri / Computational-Intelligence

Repository CI 23/24
MIT License
0 stars 0 forks source link

lab 2 review #2

Open m-depascale opened 11 months ago

m-depascale commented 11 months ago

Hello Matteo, I'm reviewing your implementation of the evolutionary strategy for creating an agent which is able to play nim.

The notes are very clear and helps in understanding your thought-flow when implementing the strategy.

I really liked how you set some conditions to choose what do to and weighted them, you can surely tune some of them to allow more precise moves, since we know how to trick nim games. Anyway, I think this is a really good implementation and by "copying" the mechanism of neurons you can achieve great results. Also, with that sigma you can achieve a good balance between exploration and exploitation, and if we just want to speed-up the process, an higher value for the sigma will be needed at the beginning.

Overall, the code is well structured and the plus strategy adopted helps in maintaining a sort of steady state, so not losing good information for future generations.

Matteo-Pietro-Pillitteri commented 11 months ago

Hi Francesco, thanks for you review!