Hello,
i’m very interested in Neural Network, Deep Learning and all this kind of stuff, and i recently was asking myself if it was possible to make an Artificial Neural Network in Defold with “genetical renforcement” (which means it learn by himself by selecting the best specimen of each generation and reproducing it with some mutation). So i created my own, it’s just a car who need to finish a race the fastest possible without hiting any wall. The network use 5 input, it’s 5 ray_cast on the front and side of the car with a length of 750pxl who react to the wall and give back the message.fraction to the network. The network have 5 hidden layers’s neurons and 2 output (turn left or right).
Visual representation of my neural network :
What the “generator” look like :
If some of you want to try it, or to modify/use it, here it is :
https://ufile.io/1kzlg (the file can only be hosted for 30 days)
To start the “generator” once you’ve build it, just left click
The best neural network of the last generation is saved here :
C:\Users\ UserName \AppData\Roaming\AI_test