Hello Alessandro,
I find your idea of characterizing the genome very clever. By leveraging nimsum as a spectrum of possible choices, your strategy, regardless of the number of wins per generation (which, however, increases significantly over generations), has great evolutionary potential, as evidenced by the collected data, where the population initially randomly targeting all nimsum values tends to concentrate in a specific range as it evolves.
I also appreciate providing two methods of parent selection because it provides a more in-depth view of the problem, highlighting aspects that often take a back seat, such as convergence speed and population diversity.
Regarding the crossover function, I appreciate the idea of having a bias that influences the result (with the check if total_fitness > 0).
The mutation is standard but efficient.
Overall, the code is well-structured and very readable. The efforts to plot all the results to make them clear are commendable, and the report is simple but effective, containing everything one needs to know and providing an excellent guide to understanding your code.
Well done, and good luck with the upcoming labs!
Hello Alessandro, I find your idea of characterizing the genome very clever. By leveraging nimsum as a spectrum of possible choices, your strategy, regardless of the number of wins per generation (which, however, increases significantly over generations), has great evolutionary potential, as evidenced by the collected data, where the population initially randomly targeting all nimsum values tends to concentrate in a specific range as it evolves. I also appreciate providing two methods of parent selection because it provides a more in-depth view of the problem, highlighting aspects that often take a back seat, such as convergence speed and population diversity. Regarding the crossover function, I appreciate the idea of having a bias that influences the result (with the check if total_fitness > 0). The mutation is standard but efficient. Overall, the code is well-structured and very readable. The efforts to plot all the results to make them clear are commendable, and the report is simple but effective, containing everything one needs to know and providing an excellent guide to understanding your code. Well done, and good luck with the upcoming labs!