Unifying learning in games and graphical models
The ever increasing use of intelligent multi-agent systems poses increasing demands upon them. One of these is the ability to reason consistently under uncertainty. This, in turn, is the dominant characteristic of probabilistic learning in graphical models which, however, lack a natural decentralise...
| Main Authors: | , , , , , |
|---|---|
| Format: | Conference item |
| Published: |
2005
|
| _version_ | 1797073547598757888 |
|---|---|
| author | Rezek, I Roberts, S Rogers, A Dash, R Jennings, N IEEE |
| author_facet | Rezek, I Roberts, S Rogers, A Dash, R Jennings, N IEEE |
| author_sort | Rezek, I |
| collection | OXFORD |
| description | The ever increasing use of intelligent multi-agent systems poses increasing demands upon them. One of these is the ability to reason consistently under uncertainty. This, in turn, is the dominant characteristic of probabilistic learning in graphical models which, however, lack a natural decentralised formulation. The ideal would, therefore, be a unifying framework which is able to combine the strengths of both multi-agent and probabilistic inference In this paper we present a unified interpretation of the inference mechanisms in games and graphical models. In particular, we view fictitious play as a method of optimising the Kullback-Leibler distance between current mixed strategies and optimal mixed strategies at Nash equilibrium. In reverse, probabilistic inference in the variational mean-field framework can be viewed as fictitious game play to learn the best strategies which explain a probabilistic graphical model. © 2005 IEEE. |
| first_indexed | 2024-03-06T23:23:43Z |
| format | Conference item |
| id | oxford-uuid:69a9522c-cb16-492c-b3d3-0820be36741a |
| institution | University of Oxford |
| last_indexed | 2024-03-06T23:23:43Z |
| publishDate | 2005 |
| record_format | dspace |
| spelling | oxford-uuid:69a9522c-cb16-492c-b3d3-0820be36741a2022-03-26T18:52:20ZUnifying learning in games and graphical modelsConference itemhttp://purl.org/coar/resource_type/c_5794uuid:69a9522c-cb16-492c-b3d3-0820be36741aSymplectic Elements at Oxford2005Rezek, IRoberts, SRogers, ADash, RJennings, NIEEEThe ever increasing use of intelligent multi-agent systems poses increasing demands upon them. One of these is the ability to reason consistently under uncertainty. This, in turn, is the dominant characteristic of probabilistic learning in graphical models which, however, lack a natural decentralised formulation. The ideal would, therefore, be a unifying framework which is able to combine the strengths of both multi-agent and probabilistic inference In this paper we present a unified interpretation of the inference mechanisms in games and graphical models. In particular, we view fictitious play as a method of optimising the Kullback-Leibler distance between current mixed strategies and optimal mixed strategies at Nash equilibrium. In reverse, probabilistic inference in the variational mean-field framework can be viewed as fictitious game play to learn the best strategies which explain a probabilistic graphical model. © 2005 IEEE. |
| spellingShingle | Rezek, I Roberts, S Rogers, A Dash, R Jennings, N IEEE Unifying learning in games and graphical models |
| title | Unifying learning in games and graphical models |
| title_full | Unifying learning in games and graphical models |
| title_fullStr | Unifying learning in games and graphical models |
| title_full_unstemmed | Unifying learning in games and graphical models |
| title_short | Unifying learning in games and graphical models |
| title_sort | unifying learning in games and graphical models |
| work_keys_str_mv | AT rezeki unifyinglearningingamesandgraphicalmodels AT robertss unifyinglearningingamesandgraphicalmodels AT rogersa unifyinglearningingamesandgraphicalmodels AT dashr unifyinglearningingamesandgraphicalmodels AT jenningsn unifyinglearningingamesandgraphicalmodels AT ieee unifyinglearningingamesandgraphicalmodels |