rasenqt
commented
9 months ago Thank you for the review. It would be cool to try your algortihm choice that cause cost decresing from n^3 to n log n for the hamming distance, and mby finding something computationally lighter for entropy metrics too. Talking about diversity, of course , our curiosity was to trying to introduce diversity inter individual and this is done by hamming distance , and then we tried to understand how diversity intra individual works, and in this case we found out that working in fitness landscape full of local maximum is a very hard task. So my question is : Is the genome structure suitable for this problem (Since in this case is hard to find a correlation between genes)?
Code Review
Code and Documentation
Hi Michelangelo and Silvio, congratulations on your work! I found your ideas really inspiring and the results are very interesting. I like how you showed the vanilla and the enhanced method one after the other to show the improvements, it really shows the power of your code.
As a personal perference I would commented a little bit more each function just to make it easier to understand for someone who is not familiar with the code. However the code is very well structured, so it was quite easy to follow. I just would preferred not to have to read the code to understand what each function does.
Algorithm
It was interesting to see the application of the concept of diversity in the context of the genetic algorithm. I think it is a very interesting approach but I still don't quite get why the less "distant" ones or the most "caothic" genomes should be the best ones. Of course the results are very good, but I would like to understand better the theory behind it.
As you stated in the report, the algorithm for the hamming sorting is really slow and I think it is the main bottleneck of the code. I think it would be interesting to see if there is a way to improve it. I found this interesting algorithm that finds the closest two point in a set of points in O(n log n) time: https://en.wikipedia.org/wiki/Closest_pair_of_points_problem. I think it could be interesting to see if it can be applied to the hamming sorting.
Results
The results are very good and the presentation is very clear. I like how you showed the results for the different methods. I don't have much to say about it other than it would be really interesting to see how the results change as the number of generations increases in a plot.
Conclusion
I think you did a great job and I really enjoyed reading your report. It would have been interesting to see the how you arrived to the final tweak of the parameters, but I understand that it would have been a lot of work.
Keep up the good work!