jLOAF: What is it? - NMAI-lab/JLOAF GitHub Wiki

jLOAF stands for java Learning by ObservAtion Framework. It was originally dreamed up and designed by Michael Floyd and Babak Esfandiairi. Its original form can be found at NMAI. Since then it has been upgraded and now lives on this repository.

jLOAF originally used Case Based Reasoning to perform learning from observation. Here is the paper that describes its structure, functionality and future improvements. It was tested in an obstacle avoidance, tetris and simulated soccer environments.

In its current form it provides a user with the ability to create an agent that learns by observing in a discrete or continuous domain using state of the art preprocessing methods and state of the art learning algorithms. It can learn state base behavior as well as reactive behavior. Please take a look at the how it works to get an overview of how the framework functions.