Potential Bug with Genetic Algorithm Function about mlrose HOT 1 CLOSED

Hi @nlachapelle ,

I had a look through the section of code that you pointed out, and I don't think there is a bug there. Here are the reasons why:

• What we want to do is to optimize the fitness function and return the state that results in the optimal fitness value and the fitness value of that state (i.e. best_state and best_fitness in the code). We are not interested in finding the population that gave rise to that state.
• When we create the new population, although we change the population at the line you pointed out, we don't change the fitness and state values of the problem.
• The fitness and state values are only updated if we find that the best child of the new population results in a better state (this is done at this section of the code):

if next_fitness > problem.get_fitness():
     problem.set_state(next_state)

• The fact that the population is overridden at each attempt to find a better state doesn't matter, since the population is just a means to an ends.

I hope that makes sense. If not, please feel free to ask questions and I'll do my best to answer them.

Genevieve.

from mlrose.