pathlearner
PathLearner is a tool which demonstrates finding a shortest path between two iven points on a grid through reinforcement learning. The tool is implemented as a Windows application and was tested on Windows NT and Windows 2000.
Some of PathLearner's features:
-
Clear visualization of various learning-task information such as learned policy, optimal policy, shortest path to destination, amount of information learned etc.
-
Variable-size grids with obstacle support.
Grids can be auto-generated, loaded from a file, saved to a file, or edited manually. -
Supported learning algorithms: Q and TD (both single step).
Parameters are user-configurable. -
Supported selection strategies: Random, Epsilon-Greedy, K-Probability (parameters are configurable).
-
Experiment summary can be saved as text file.
-
Written completely in C++ for high performance.
View some screenshots, download the application, or the sources.
For any other issue please contact me.