“高屋建瓴AI公开课”项目由中国人民大学高瓴人工智能学院发起，旨在扩大人工智能学科影响力、提升学科发展水准。公开课项目命名为“高屋建瓴”，寓意在高瓴人工智能学院的平台上，汇聚高端人才，发出人工智能研究方向高瞻远瞩的声音。
讲座主题：
DecisionMakingandReinforcementLearning
腾讯线上会议ID：742-740-515
活动安排：
16:00-16:10主持人介绍
16:10-16:50讲座
16:50-17:20Q&A
讲座摘要
ArtificialIntelligence(AI)aimsatcreatingacomputerisedsystemcapableofacquiringandapplyingknowledgeandskillsasaresultofexperience.BeingabletoperceivetheworldandsubsequentlyinteractingwithitarethetwopillarsofintelligencethatshouldbemaintainedinanyAIsystem.
Inscientificresearch,itiswidelyacceptedthattheformercorrespondstopatternrecognition,whilethelatterisrelatedtomachinedecisionmaking.Withthedecadesofeffortstosupervisedandunsupervisedlearning,manypatternrecognitionproblemshavebeenwellexploredandalotofsuccessfulstoriescanbefoundrangingfromspeechrecognitionandmachinetranslationtovisualobjectdetectionandrecognition.Bycontrast,machinedecisionmakingisstillatitsinfantstageduetolackofunderstandingofitsintrinsiccomplexityofthedecisionspace,andmoreresearcheffortisrequiredtodriveprogressinthisfieldforward.
Inaddition,multi-agentlearningarisesinavarietyofdomainswhereintelligentagentsinteractnotonlywiththe(unknown)environmentbutalsowitheachother.Ithasanincreasingnumberofapplicationsrangingfromcontrollingagroupofautonomousvehicles/robots/dronestocoordinatingcollaborativebotsinproductionlines,optimisingdistributedsensornetworks/traffic,andmachinebiddingincompetitivee-commerce,searchandinformationretrievalandfinancialmarkets.
Inthistalk,Ishallprovideanup-to-dateintroductiononthetechniqueandalgorithmsofmachinedecisionmakingandmulti-agentAI,withafocusoncompetition,collaboration,andcommunicationsamongintelligentagents.Thestudiesinbothgametheoryandmachinelearningwillbeexaminedinaunifiedtreatment.Ishallalsosampleourrecentworkonthesubjectincludingmean-fieldmultiagentreinforcementlearning,generativeadversarialnets,hierarchicalreinforcementlearningandtheirapplicationsininformationretrievalandsearch.
主讲嘉宾
Prof.JunWang,UCL
JunWangisProfessorattheComputerSciencedepartment,UniversityCollegeLondon.Prof.JunWang'smainresearchinterestsareintheareasofAIandintelligentsystems,covering(multiagent)reinforcementlearning,deepgenerativemodels,andtheirdiverseapplicationsoninformationretrieval,recommendersystemsandpersonalization,datamining,smartcities,botplanning,andcomputationaladvertising.
Histeamwonthefirstglobalreal-timebiddingalgorithmcontestwith80+participantsworldwide.Junhaspublishedover200researchpapersandisawinnerofmultipleBestPaperawards.HewasarecipientoftheBeyondSearch---SemanticComputingandInternetEconomicsawardbyMicrosoftResearchandalsoreceivedYahoo!FREPFacultyaward.HehasservedasanAreaChairinACMCIKMandACMSIGIR.Hisrecentserviceincludesco-chairofArtificialIntelligence,Semantics,andDialoginACMSIGIR2018.
邀请人
陈旭高瓴人工智能准聘助理教授
陈旭博士毕业于清华大学，博士期间曾在佐治亚理工学院进行交流访问，博士毕业后曾在英国伦敦大学学院担任博士后研究员。于2020年加入中国人民大学，任助理教授。其主要研究方向为推荐系统，强化学习，因果推断等。曾在SIGIR、TOIS、WWW、WSDM、CIKM、AAAI等信息检索领域顶级会议和期刊发表论文40余篇。据GoogleScholar统计，已发表论文共计被引用2000余次。曾获得TheWebConference2018最佳论文提名奖、AIRS2017最佳论文奖。陈旭曾担任SIGIR、WWW、IJCAI、AAAI、CIKM等会议的程序委员会委员，以及TOIS、TKDE、TIST、JMLR等杂志的审稿人。