Skip navigation
Skip navigation

Conditional random fields for multi-agent reinforcement learning

Zhang, Xinhua; Aberdeen, Douglas; Vishwanathan, S

Description

Conditional random fields (CRFs) are graphical models for modeling the probability of labels given the observations. They have traditionally been trained with using a set of observation and label pairs. Underlying all CRFs is the assumption that, conditioned on the training data, the labels are independent and identically distributed (iid). In this paper we explore the use of CRFs in a class of temporal learning algorithms, namely policy-gradient reinforcement learning (RL). Now the labels are...[Show more]

dc.contributor.authorZhang, Xinhua
dc.contributor.authorAberdeen, Douglas
dc.contributor.authorVishwanathan, S
dc.date.accessioned2015-12-10T21:57:46Z
dc.identifier.isbn9781595937933
dc.identifier.urihttp://hdl.handle.net/1885/39922
dc.description.abstractConditional random fields (CRFs) are graphical models for modeling the probability of labels given the observations. They have traditionally been trained with using a set of observation and label pairs. Underlying all CRFs is the assumption that, conditioned on the training data, the labels are independent and identically distributed (iid). In this paper we explore the use of CRFs in a class of temporal learning algorithms, namely policy-gradient reinforcement learning (RL). Now the labels are no longer iid. They are actions that update the environment and affect the next observation. From an RL point of view, CRFs provide a natural way to model joint actions in a decentralized Markov decision process. They define how agents can communicate with each other to choose the optimal joint action. Our experiments include a synthetic network alignment problem, a distributed sensor network, and road traffic control; clearly outperforming RL methods which do not model the proper joint policy.
dc.publisherAssociation for Computing Machinery Inc (ACM)
dc.relation.ispartofMachine Learning
dc.relation.isversionof1st Edition
dc.subjectKeywords: Data reduction; Gradient methods; Mathematical models; Multi agent systems; Probability; Random number generation; Markov decision processes; Multi-agent reinforcement learning; Random fields; Reinforcement learning
dc.titleConditional random fields for multi-agent reinforcement learning
dc.typeBook chapter
local.description.notesImported from ARIES
dc.date.issued2007
local.identifier.absfor080109 - Pattern Recognition and Data Mining
local.identifier.ariespublicationu8803936xPUB185
local.type.statusPublished Version
local.contributor.affiliationZhang, Xinhua, College of Engineering and Computer Science, ANU
local.contributor.affiliationAberdeen, Douglas, College of Engineering and Computer Science, ANU
local.contributor.affiliationVishwanathan, S, College of Engineering and Computer Science, ANU
local.description.embargo2037-12-31
local.bibliographicCitation.startpage1143
local.bibliographicCitation.lastpage1150
local.identifier.doi10.1145/1273496.1273640
dc.date.updated2015-12-09T07:48:10Z
local.bibliographicCitation.placeofpublicationUSA
local.identifier.scopusID2-s2.0-34547980892
CollectionsANU Research Publications

Download

File Description SizeFormat Image
01_Zhang_Conditional_random_fields_for_2007.pdf140.67 kBAdobe PDF    Request a copy
02_Zhang_Conditional_random_fields_for_2007.pdf62.77 kBAdobe PDF    Request a copy
03_Zhang_Conditional_random_fields_for_2007.pdf292.9 kBAdobe PDF    Request a copy
04_Zhang_Conditional_random_fields_for_2007.pdf66.84 kBAdobe PDF    Request a copy
05_Zhang_Conditional_random_fields_for_2007.pdf744.12 kBAdobe PDF    Request a copy


Items in Open Research are protected by copyright, with all rights reserved, unless otherwise indicated.

Updated:  19 May 2020/ Responsible Officer:  University Librarian/ Page Contact:  Library Systems & Web Coordinator