duaraghav8 / genetic Goto Github PK

By the time the 6th generation rises, 77.8% of the words has already been formed (only 2 out of 9 alphabets left). However, the fitness simply doesn't improve after this (as can bee seen in output). The population list gets exhausted despite providing it with 90,000 unique combinations.

1 possibility could be that I haven't used Mutation in the script yet. Another could be the wrong crossover mask applied.
Still working on it...

UPDATE: The evolve_2.py performs better than evolve and reaches 88% accuracy in the 6th generation (generates the word "zvolution"). The only modification I made was the crossover mask. In evolve_2, I used a randomly generated mask in contrast to the evolve's single mask (111100000).

This highlights the importance of choosing the correct Crossover Mask.

UPDATE: It done. You can see in output_3 that the fitness is 100% and the word has correctly been guessed. I added on to script_2 by adding up Mutation to about 10% of the new generation. That did the job.
Looks like I underestimated the significance of crossover mask and mutation.