June 10, 2004

This is a simple model of an artificial intelligence learning program that plays Tetris. It's a bit choppy and hackish around the edges and may require a lot of work to get it to a satisfactory state. It basically works by heavy use of a critic function. If the critic determines that a specific decision tree answer is absolutely silly, it will find what it thinks is the best possible answer by means of brute force.

It has two modes: one that depends solely on the critic for a "control" and the other is a watch mode where a human can play Tetris and it will basically save the data to the tree.

There is no GUI and the interactive and non-interactive modes are console based. It's written in c++, is heavily dependent on the STL, and interfaces with c5.0 on the local system.

The project can be found on GitHub here.

Further Reading

Related Posts