{"id":1051,"date":"2018-01-11T13:14:12","date_gmt":"2018-01-11T12:14:12","guid":{"rendered":"http:\/\/blogs.gm.fh-koeln.de\/ciop\/?page_id=1051"},"modified":"2025-01-20T09:06:59","modified_gmt":"2025-01-20T08:06:59","slug":"gbg-games-learning","status":"publish","type":"page","link":"https:\/\/blogs.gm.fh-koeln.de\/ciop\/research\/gbg-games-learning\/","title":{"rendered":"Games & Learning"},"content":{"rendered":"

\"c4board\"<\/p>\n

+++ Update 11\/2024: FutureWorkGBG<\/strong>: Project themes in GBG for students<\/a> (PDF or DOCX) +++<\/p>\n

(see also: GBG: General Board Games<\/a> in CIOP )<\/span><\/p>\n

Games are interesting in computer science concerning the question whether a computer can learn <\/strong><\/span>the game strategies just from self-play, without explicitly programming the tactics or performing exhaustive search. This is a branch of artificial intelligence (AI)<\/strong><\/span>. Many game learning approaches are based on reinforcement learning<\/strong><\/span>, namely TD-learning.
\n<\/span><\/p>\n

In our research group we have studied extensively the game Connect-4<\/strong><\/span> (\"Four-in-a-Row\"). We were able to develop an agent which learns Connect-4 nearly perfectly just from self-play. Our Java-based Connect-4 Game Playing Framework (C4GPF)<\/a> is open-source<\/em>\u00a0for interested researchers<\/span>. Read more...<\/a><\/span> <\/span><\/p>\n

 <\/p>\n

In January'2018 we released \"Gameboard<\/a>GBG<\/strong><\/span>, the G<\/strong><\/span>eneral<\/strong> B<\/strong><\/span>oard<\/strong> G<\/strong><\/span>ame<\/strong> playing and learning framework<\/strong> to the research community as another open-source<\/em> project.<\/span><\/span><\/p>\n

https:\/\/github.com\/WolfgangKonen\/GBG<\/a><\/p>\n

Read more about GBG ...<\/a><\/p>\n

The long-term goal of our research group is it to transfer these learning strategies to many other games (dots-and-boxes, go, Poker, checkers, Abalone, Sim, Othello, ...). The project is related to the research field known as General Game Playing (GGP)<\/strong><\/span>. The aim of GGP is it to develop agents which are able to learn a great variety of games. <\/span><\/p>\n

 <\/p>\n

People<\/h2>\n

Wolfgang Konen<\/a>, Markus Thill<\/a>, Samineh Bagheri<\/a>, Johannes Kutsch, Kevin Galitzki, Felix Barsnick, Julian C\u00f6ln, Yannick Dittmar, Johannes Scheiermann<\/p>\n

Publications<\/h2>\n

<\/a><\/form>

2023<\/h3>

Konen, Wolfgang<\/p>

Towards Learning Rubik's Cube with N-tuple-based Reinforcement Learning<\/a> Artikel<\/span> <\/p>

In: <\/span>arXiv preprint arXiv:2301.12167, <\/span>2023<\/span>.<\/p>

Links<\/a><\/span> | BibTeX<\/a><\/span><\/p>

@article{Konen2023,
\r\ntitle = {Towards Learning Rubik's Cube with N-tuple-based Reinforcement Learning},
\r\nauthor = {Wolfgang Konen},
\r\nurl = {https:\/\/arxiv.org\/abs\/2301.12167},
\r\nyear  = {2023},
\r\ndate = {2023-01-01},
\r\nurldate = {2023-01-01},
\r\njournal = {arXiv preprint arXiv:2301.12167},
\r\nkeywords = {},
\r\npubstate = {published},
\r\ntppubtype = {article}
\r\n}
\r\n<\/pre><\/div>
Schlie\u00dfen<\/a><\/p><\/div>