 |
Reinforcement Learning and
Artificial
Intelligence (RLAI)
|
RLAI
Software Development Resources Page
|
The ambition of this
web
page is to provide information and resources for software currently
under development by
the
RLAI group at the University of Alberta. This page links to current development
issues and code, describes some future directions, and also contains
links
to resources for languages and operating systems used in our
development. Tools listed on this page are the very latest versions,
with new capabilities, but incompletely tested.
This page is intended to be a meeting place for
discussion of code level issues and the updating of pre-beta test code
for the various RLAI software projects. It is mostly intended for the
members of the group doing the development, but may also be of interest
to users eager for the latest version and willing to help test out
potentially buggy code.
Latest Version of Tools (Beta at best, possibly even Alpha
versions)
- RL
toolkit - a package of reinforcement learning tools and demos in
python. The
toolkit may be useful to anyone
trying to learn, teach or use reinforcement learning. The tools are
suitable
for
a range of users, from new users who have never used RL before, to very
experienced users. The latest code is available in a CVS
repository. See here for
instructions on how to use this to get and change code. Following are
development pages for some aspects of the toolkit. Code and new
documentation is available on these pages as well.
- Gridworld
Demo - GUI and nonGUI versions of an agent making its way through a
gridworld to a goal
- G - the 2-d graphics module
- Tiles - tile coding routines
- RLinterface - the standard for
connecting agents and environments
- The Open Pages system - the easiest
web page editor/manager around!
- ... please add
Tools under development by the RLAI group
- A reinforcement learning competition, to highlight the best
algorithms for solving RL problems of various sorts. Early planning
stages. To be based on the RLI.
See here for
the NIPS workshop.
- A standard set of RL task and agent implementations, to
facilitate comparisons. To be based on the RLI.
This has long been
talked about, but never achieved. There is work going on it at the
moment. Its progress can be tracked here.
- Educational software, for teaching RL.
- ... please add
Future Directions
- Web-accessable demonstrations of RL in action.
- Adding sound output to the existing RL demos.
- ... please add
Information on Tools used by the RLAI Group
Language Support
OS Support