Reinforcement Learning

programme

Licenses

Source code

Soumis par

Crédits

IMAGINARY

Contributeurs

Soutenu(e) par

Carl Zeiss Foundation

An AI that uses Reinforcement Learning learns through trial and error by directly interacting with its environment. It’s similar to how humans and animals learn.

The exhibit offers three different panels in which a robot tries to find its way through a maze. You can help it learn through placing rewards and also change its Exploration:Exploitation ratio. Create your own mazes and the robot will slowly learn to find the exit and to avoid dangerous areas. Experiment with leaving bonus rewards on their way to reinforce certain paths.

 

Fichier principal

Répertoire

Fichiers