Question: Attempting to run the Genetic Algorithm from mlrose on the Four Peaks problem, but it converges very fast and not sure if this is normal.
Attempting to run the Genetic Algorithm from mlrose on the Four Peaks problem, but it converges very fast and not sure if this is normal. I've used 7 different seeds and a population size of 600, and 30 iterations for each.
fitness_fn = mlrose.FourPeaks(t_pct=.1) problem = mlrose.DiscreteOpt(length=600, fitness_fn=fitness_fn, maximize=True, max_val=2) mlrose.genetic_alg(problem, pop_size=int(600), mutation_prob=0.4, max_iters=int(max_iter), random_state=seed)
1 9.0 10.0 13.0 11.0 13.0 10.0 9.0 2 11.0 12.0 115.0 11.0 13.0 16.0 9.0 3 11.0 16.0 115.0 13.0 17.0 112.0 9.0 4 13.0 16.0 119.0 13.0 17.0 112.0 9.0 5 22.0 24.0 120.0 14.0 17.0 114.0 12.0 6 22.0 24.0 120.0 20.0 17.0 116.0 12.0 7 122.0 24.0 120.0 113.0 17.0 116.0 12.0 8 122.0 24.0 120.0 114.0 17.0 117.0 12.0 9 122.0 24.0 120.0 120.0 19.0 117.0 112.0 10 122.0 112.0 120.0 123.0 19.0 117.0 112.0 11 124.0 116.0 120.0 123.0 20.0 117.0 114.0 12 124.0 121.0 120.0 123.0 20.0 117.0 114.0 13 124.0 121.0 120.0 123.0 20.0 117.0 114.0 14 125.0 121.0 120.0 123.0 20.0 117.0 115.0 15 126.0 121.0 120.0 123.0 20.0 119.0 115.0 16 127.0 121.0 120.0 123.0 21.0 119.0 115.0 17 127.0 121.0 120.0 123.0 21.0 119.0 115.0 18 127.0 121.0 120.0 123.0 21.0 119.0 117.0 19 127.0 121.0 120.0 123.0 21.0 119.0 117.0 20 127.0 121.0 120.0 123.0 114.0 119.0 122.0 21 127.0 121.0 120.0 123.0 117.0 119.0 122.0 22 127.0 121.0 120.0 123.0 119.0 119.0 122.0 23 127.0 121.0 120.0 123.0 120.0 120.0 122.0 24 127.0 121.0 120.0 123.0 121.0 123.0 122.0 25 127.0 121.0 120.0 123.0 121.0 123.0 122.0 26 127.0 121.0 120.0 123.0 121.0 123.0 122.0 27 127.0 121.0 120.0 123.0 121.0 123.0 122.0 28 127.0 121.0 120.0 123.0 122.0 123.0 122.0 29 127.0 121.0 120.0 123.0 123.0 123.0 122.0 30 127.0 121.0 120.0 123.0 123.0 123.0 122.0
The question is, why is it converging so fast and what can I do to improve?
Step by Step Solution
There are 3 Steps involved in it
Get step-by-step solutions from verified subject matter experts
