Project 1 – NNP – Neural Network Pong

I intend to create a Flash version of the classic game Pong.

Then implement a MLP (Multi-Layered-Perceptron) in AS3 (Actionscript 3).

Players will then start with a ball, a wall and a paddle.  They will hit the ball against the wall with the paddle in the sense of a 1 player Pong game.  After a while enough data will be gathered to train the NN (neural network) to the players style.

Training of the NN will take place and then the player will be able to play against the trained NN.

Firstly I hope it works, it should be able to move the paddle to react to the ball.  Obviously it will only be as good as the player who trains it.  Eventually I want it to be able to possibly train as play goes on.

I hope the AS3 implementation will be fast enough, as Actionscript is not choosen for its execution speed, (nor its debugging for that matter).

I also intend to create a visual representation of the NN in the form of the nodes and connecting lines, the width determind by the weight of each connection.  You should be able to watch the NN develop and update its weights.

Posted on September 17, 2008 at 8:18 am by Jordan Carter · Permalink
In: Artifical Intelligence · Tagged with: , , ,

Leave a Reply