New Hanabi Paper
parent
4a51abff5a
commit
9570cc292a
|
|
@ -166,6 +166,7 @@ If you aren't able to access any paper on this list, please [try using Sci-Hub](
|
|||
- [Emergence of Cooperative Impression With Self-Estimation, Thinking Time, and Concordance of Risk Sensitivity in Playing Hanabi](https://www.frontiersin.org/articles/10.3389/frobt.2021.658348/full) (journalArticle)
|
||||
- []() (conferencePaper)
|
||||
- [K-level Reasoning for Zero-Shot Coordination in Hanabi](https://papers.neurips.cc/paper/2021/hash/4547dff5fd7604f18c8ee32cf3da41d7-Abstract.html) (conferencePaper)
|
||||
- [Is Vanilla Policy Gradient Overlooked? Analyzing Deep Reinforcement Learning for Hanabi](http://arxiv.org/abs/2203.11656) (journalArticle)
|
||||
|
||||
# Hearthstone
|
||||
- [Mapping Hearthstone Deck Spaces through MAP-Elites with Sliding Boundaries](http://arxiv.org/abs/1904.10656) (journalArticle)
|
||||
|
|
|
|||
|
|
@ -8907,6 +8907,67 @@ guaranteed decent high score. The algorithm got a lowest score of 79 and a
|
|||
<z:linkMode>1</z:linkMode>
|
||||
<link:type>application/pdf</link:type>
|
||||
</z:Attachment>
|
||||
<bib:Article rdf:about="http://arxiv.org/abs/2203.11656">
|
||||
<z:itemType>journalArticle</z:itemType>
|
||||
<dcterms:isPartOf>
|
||||
<bib:Journal><dc:title>arXiv:2203.11656 [cs]</dc:title></bib:Journal>
|
||||
</dcterms:isPartOf>
|
||||
<bib:authors>
|
||||
<rdf:Seq>
|
||||
<rdf:li>
|
||||
<foaf:Person>
|
||||
<foaf:surname>Grooten</foaf:surname>
|
||||
<foaf:givenName>Bram</foaf:givenName>
|
||||
</foaf:Person>
|
||||
</rdf:li>
|
||||
<rdf:li>
|
||||
<foaf:Person>
|
||||
<foaf:surname>Wemmenhove</foaf:surname>
|
||||
<foaf:givenName>Jelle</foaf:givenName>
|
||||
</foaf:Person>
|
||||
</rdf:li>
|
||||
<rdf:li>
|
||||
<foaf:Person>
|
||||
<foaf:surname>Poot</foaf:surname>
|
||||
<foaf:givenName>Maurice</foaf:givenName>
|
||||
</foaf:Person>
|
||||
</rdf:li>
|
||||
<rdf:li>
|
||||
<foaf:Person>
|
||||
<foaf:surname>Portegies</foaf:surname>
|
||||
<foaf:givenName>Jim</foaf:givenName>
|
||||
</foaf:Person>
|
||||
</rdf:li>
|
||||
</rdf:Seq>
|
||||
</bib:authors>
|
||||
<dc:subject>
|
||||
<z:AutomaticTag>
|
||||
<rdf:value>Computer Science - Artificial Intelligence</rdf:value>
|
||||
</z:AutomaticTag>
|
||||
</dc:subject>
|
||||
<dc:subject>
|
||||
<z:AutomaticTag>
|
||||
<rdf:value>Computer Science - Machine Learning</rdf:value>
|
||||
</z:AutomaticTag>
|
||||
</dc:subject>
|
||||
<dc:subject>
|
||||
<z:AutomaticTag>
|
||||
<rdf:value>Computer Science - Multiagent Systems</rdf:value>
|
||||
</z:AutomaticTag>
|
||||
</dc:subject>
|
||||
<dc:title>Is Vanilla Policy Gradient Overlooked? Analyzing Deep Reinforcement Learning for Hanabi</dc:title>
|
||||
<dcterms:abstract>In pursuit of enhanced multi-agent collaboration, we analyze several on-policy deep reinforcement learning algorithms in the recently published Hanabi benchmark. Our research suggests a perhaps counter-intuitive finding, where Proximal Policy Optimization (PPO) is outperformed by Vanilla Policy Gradient over multiple random seeds in a simplified environment of the multi-agent cooperative card game. In our analysis of this behavior we look into Hanabi-specific metrics and hypothesize a reason for PPO's plateau. In addition, we provide proofs for the maximum length of a perfect game (71 turns) and any game (89 turns). Our code can be found at: https://github.com/bramgrooten/DeepRL-for-Hanabi</dcterms:abstract>
|
||||
<dc:date>2022-03-22</dc:date>
|
||||
<z:shortTitle>Is Vanilla Policy Gradient Overlooked?</z:shortTitle>
|
||||
<z:libraryCatalog>arXiv.org</z:libraryCatalog>
|
||||
<dc:identifier>
|
||||
<dcterms:URI>
|
||||
<rdf:value>http://arxiv.org/abs/2203.11656</rdf:value>
|
||||
</dcterms:URI>
|
||||
</dc:identifier>
|
||||
<dcterms:dateSubmitted>2022-03-26 04:22:52</dcterms:dateSubmitted>
|
||||
<dc:description>arXiv: 2203.11656</dc:description>
|
||||
</bib:Article>
|
||||
<z:Collection rdf:about="#collection_6">
|
||||
<dc:title>2048</dc:title>
|
||||
<dcterms:hasPart rdf:resource="https://doi.org/10.1007%2F978-3-319-50935-8_8"/>
|
||||
|
|
@ -9016,6 +9077,7 @@ guaranteed decent high score. The algorithm got a lowest score of 79 and a
|
|||
<dcterms:hasPart rdf:resource="https://www.frontiersin.org/articles/10.3389/frobt.2021.658348/full"/>
|
||||
<dcterms:hasPart rdf:resource="#item_566"/>
|
||||
<dcterms:hasPart rdf:resource="https://papers.neurips.cc/paper/2021/hash/4547dff5fd7604f18c8ee32cf3da41d7-Abstract.html"/>
|
||||
<dcterms:hasPart rdf:resource="http://arxiv.org/abs/2203.11656"/>
|
||||
</z:Collection>
|
||||
<z:Collection rdf:about="#collection_55">
|
||||
<dc:title>Hearthstone</dc:title>
|
||||
|
|
|
|||
Loading…
Reference in New Issue