Rating is available when the video has been rented.
This feature is not available right now. Please try again later.
Published on Nov 5, 2009
An expert (near-optimal planner) trains the computer to play Mario Bros on randomly generated levels of varying difficulty (simulator from the recent Mario Bros AI competition http://julian.togelius.com/mariocompe... Binary features are extracted from screenshot images and corresponding actions (left,right,jump,speed) are recorded at 6 Hz. Neural networks are used to learn the action based off input features. Results for SMILe are obtained using a learning rate alpha=0.1.