project PongGame with SNN V12 (2024)

This is a demonstration of an AI playing Pong Game. Learning is Hebbian-like. This version works with Auto-Suggest Key. This type of learning resembles two friends teaching each other a computer game, where one friend keeps telling the other which key to press. First, the network is pre-trained to teach the AI ​​which key input corresponds to which other key output. There is no error backpropagation (like I told only Hebbian), there's only association of a random chosen output to which a certain input is fed. Limitation is the reaction time of the network, this is 1.6 seconds. It takes 1.6 seconds for the network to recognize an input pattern. This is also visible as the ball is sometimes missed by the AI. For the rest the actions are reasonable, hope you enjoy my video.