RLAI open page

	Reinforcement Learning and Artificial Intelligence (RLAI)
	Tool team meeting

The ambition of this page is to plan and discuss the development of software tools to support RLAI research at the University of Alberta. In particular, to identify the highest priority objectives and track progress toward them. This page is expected to be edited only by members of the UAlberta tool team. All others, particularly possible users of the tools, are strongly encouraged to make suggestions, comments, requests, recommentations, complaints, questions, etc., by extending this page or editing the extensions portion of this page.

Back to rlai.net

Definitions: A software tool is a software component that performs a clearly identifiable function and can be used as part of a larger system or development process. An ideal tool is one that performs a single function particularly well (and thus adds value), has a clear and simple interface (so that it can be used in multiple ways), and is small and readable (so that it is easy to understand and use with confidence). Larger systems, such as development environments or programming languages, can also be considered tools to the extent that they have some generality and can be used in support of a larger effort.

The tool team will meet once a week, with a presentation every 4th week at the RLAI group meeting. Time and place to be determined.

The UAlberta Tool Team

Mark Lee, lead programmer
Anna Koop, part-time magician
Rich Sutton, RLAI lab director

Contributions and participation by others is more than welcome. At some point the scope of this effort is expected to extend beyond UAlberta.

Mission Statement

The primary mission of the UAlberta tool team is to support and coordinate software development by the RLAI research team (graduate students, postdoctoral fellows, and faculty) at the University of Alberta. A secondary mission is to support and coordinate the transfer of RLAI-related software to industry and other academic institutions in a professional manner.

In practice, this means creating, finding, refining, or interfacing to

Rapid-prototyping development environments
Tools for system integration, such as interfacing high- and low-level programming languages
Tools that achieve the same functions in both high- and low-level programming languages, or that can be used across them.
Professional external documentation (e.g., web pages) such that supported tools can be easily obtained and used; accessable examples of their use
Tools for version control of software components
Web tools to support and encourage collaboration among RLAI researchers

Perhaps the greatest challenge to tool development is knowing what will actually in the end be useful to researchers. The researchers themselves are often unable to anticipate their needs with any accuracy. Developers must balance responding to newly recognized needs against working long enough on a particular need to achieve a quality product of lasting utility. In practice this means frequent interaction with researchers, interviewing them and soliciting their suggestions, comments, and feedback and then making judicious choices of a limited set of a high priority needs that can be addressed with the minimal long-term committment.

Table of Tools, Priotity, Status (in progress/complete, alpha, beta, tested, polished, reviewed, frozen, examples, documented), Completion date, Responsible person

online version of RLAI book

Tiles in Python, C++, Lisp

Standard for RL benchmarks/interfaces and associated examples

Open pages 1.0, with archiving, de-subscribing, and semi-private pages

Gridworld example

G in Python, C++, Lisp; finalizing tk version; Quartz version; object-based redesign

Graph in Python, C++, Lisp
Other graphics and data analysis tools

Function approximators, FA visualizers

Extend this Page How to edit Style Subscribe Notify Suggest Help This open web page hosted at the University of Alberta. Terms of use 2057/0