A simple game of dodge the enemy, with a bit of machine Learning Genetic Algorithm to make it better

UPDATE 1/21/2021: -Fixed the crossover function, this previously had a 31.25% chance to get a parents connection and now 98.75% chance, this was a problem since it was deleting all valued connections

-Bigger fix is that after a few more projects (Car Driving and a Cell Evolution) one error I figured out that due to the crossover function there was a point where I set the childs connections for a certain node equal to the parents and that was in memory not creating a copy, but linking them so if you mutated the child it mutated the parent. This is changed to make a copy and now gets 6-8 average score with no hidden layers in like 20-30 generations and could probably go higher with more time/layers. Saw it get to 80+ score and I have the game stop at 200 score, since if you get there you pretty much can't lose so it's getting pretty close

Will keep coming back and updating this more if I find more improvements in my algorithim

All algorithims are self wrote, but the mutation functions exact values was influenced by CodeBullet's dinoGame as a lot of my machine learning ideas stem from him and Carykh (plus, I have no idea good values for mutations, I originally wrote another function entirely and was never evolving basically).