This is an implementation of the Logistical Regression algorithm applied to a simple target shooting game. A target rockets across the sky and the goal is for the learning algorithm to train to accurately shoot the target down.
If you are curious about the algorithms used in this game, visit www.ml-class.org
The game is using a Java Applet to run the logistical regression, so please make sure you have a recent version of Java installed and that you do not have any browser plugins that prevent your browser from executing Java Applets.
You will need to provide some training examples for the algorithms to learn from. Start by trying to shoot the targets yourself. Once you have collected a number of training samples (ex. 50), you can try engaging the Auto-Pilot. If you notice that that Auto-Pilot is not very accurate or too hesitant, try collecting some more training data. You can use the "Random Shooter" button to make the cannon shoot "randomly" and try to automatically collect training data.
Hint: a miss, thought might be damaging to the shooter's ego, is still very valuable information for the learning algorithm. So don't try to be perfect.
Tweet Start Playing