search

wagenaartje / neataptic

:rocket: Blazing fast neuro-evolution & backpropagation for the browser and Node.js
https://wagenaartje.github.io/neataptic/
Other
1.18k stars 278 forks source link

How to define output for evolve without target in game. #152

Closed AZE38 closed 5 years ago

AZE38 commented 5 years ago

hi, i want to train a network with evolve, i dont have output target, i just want to take the "better" network ,with my "condtion" in game -> "create-news and mutate" from this network ! I can do this ? if yes, how ? ^^ ( sry for my english..)

thx !

AZE38 commented 5 years ago

beacause all of i have seen, i have to define

var trainingSet = [ { input: [0,0], output: [0] }, { input: [0,1], output: [1] }, { input: [1,0], output: [1] }, { input: [1,1], output: [0] } ];

But i have juste one input [0...,i,.....n], who change in each "FPS" , and no "output"... so i want to "decomposed" the training, and select each "better" for my next generation.

AZE38 commented 5 years ago

hoo..i just must to do my own selection ->crossover->mutation .... from all of my starting network ? and stop when i m happy.

AZE38 commented 5 years ago

sry for useless post ^^

rhysstubbs commented 5 years ago

@AZE38 please close the issue if this is resolved.