Lab 9 peer review - Matteo Pietro Pillitteri s314404

[Matteo-Pietro-Pillitteri]

Hi Luca, i have some advice for you. I hope that this review can help you.

• I think that is very important to write a README. It could include some clarifications about:

  • the main choses that you made in your work (e.g what algorithm you implemented)
  • some motivation about the hyperparameters that you chose (e.g mutation probability, offspring size..)
  • everything that is relevant to well understanding your code

• I immediately noticed the of_over. I think this is a form of "cheating" because you shouldn't know that fitness increases if the number of 1's in the genome increases. So, don't use it to evaluate the performance of your algorithm.

• I see that your results are obtained using a one_cut_crossover. You can try using a two_cut _crossover to increase diversity during recombination and you may get slightly better results with just this small variation. I'll leave you my implementation for the 2-cut:

def two_cut_xover(ind1: Individual, ind2: Individual) -> Individual:

while True:
    cut_point_one = randint(0, GENOME_LENGTH -1)
    cut_point_two = randint(0, GENOME_LENGTH -1)
    if cut_point_one != cut_point_two:
        break


if cut_point_one > cut_point_two:
    temp = cut_point_one
    cut_point_one = cut_point_two
    cut_point_two = temp

offspring = Individual(
    fitness=None,
    genotype = ind1.genotype[:cut_point_one]
    + ind2.genotype[cut_point_one:cut_point_two]
    + ind1.genotype[cut_point_two:]
)

return offspring

• For improving you results I can advice you to try different mutation operators such as scramble_mutation, random_resetting, reverse_mutation, gaussian_mutation, swap_mutation.
  Recommended reading: https://www.geeksforgeeks.org/mutation-algorithms-for-string-manipulation-ga/
  Some other comments:
• I really liked your implementation of uniform crossover

[FedeBucce]

Thank you for your review!