Message Boards Message Boards

Use machine learning for that the computer learn how to play a simple game?

GROUPS:

(Sorry if there are mistakes, I'm not English) Hello everyone, I'm a student which come in a stage in the British Wolfram, and we have to make a little project on the 2 last days. Me and my friend we had the idea to create a very little game : a red point at the centre of a square need to move and eat other green points, and we wanted to build a neural network for that the computer learn how to play, the only problem is that, because of the recent update to Mathematica 11, we don't really know how to use the functions for build a neural network. If someone can help us to understand what functions we should use and in which order, it will help us a lot. Thank you in advance.

POSTED BY: Mathieu DEGRE
Answer
23 days ago

You would probably be better off using genetic algorithms for this. Neural networks depend on a large set of training data, which you probably do not have. As I understand it, genetic algorithms start from zero knowledge and a set of algorithms and randomly modify them over time to achieve a goal.

You will need to consult a reference for genetic algorithms, since at the moment, they are 'roll your own' in Wolfram language.

POSTED BY: George Woodrow III
Answer
21 days ago

Group Abstract Group Abstract