ICML2019-深度强化学习文章汇总
发布时间：2019-05-10 21:47:25 来源：自考网深度强化学习-Report
来源：icml2019conference
编辑：DeepRL
强化学习是一种通用的学习、预测和决策范式。RL为顺序决策问题提供了解决方法，并将其转化为顺序决策问题。RL与优化、统计学、博弈论、因果推理、序贯实验等有着深刻的联系，与近似动态规划和最优控制有着很大的重叠，在科学、工程和艺术领域有着广泛的应用。
RL最近在学术界取得了稳定的进展，如Atari游戏、AlphaGo、VisuoMotor机器人政策。RL也被应用于现实场景，如推荐系统和神经架构搜索。请参阅有关RL应用程序的最新集合。希望RL系统能够在现实世界中工作，并具有实际的好处。然而，RL存在着许多问题，如泛化、样本效率、勘探与开发困境等。因此，RL远未被广泛部署。对于RL社区来说，常见的、关键的和紧迫的问题是：RL是否有广泛的部署问题是什么如何解决这些问题
在国际会议上的机器学习（ICML）是一个国际学术会议上机器学习。它是机器学习和人工智能研究中高影响力的两个主要会议之一。每年的ICML中都有大量的关于强化学习的文章,其中2019总共接收强化学习论文46篇（已经是很高比例了，快接近10%），下面是本次会议文章的总结，文章pdf版本汇总下载链接见文章末尾。
方法类文章
- EfficientOff-PolicyMeta-ReinforcementLearningviaProbabilisticContextVariables
- BayesianActionDecoderforDeepMulti-AgentReinforcementLearning
- QuantifyingGeneralizationinReinforcementLearning
- PolicyCertificates:TowardsAccountableReinforcementLearning
- NeuralLogicReinforcementLearning
- ProbabilityFunctionalDescent:AUnifyingPerspectiveonGANs,VariationalInference,andReinforcementLearning
- Few-ShotIntentInferenceviaMeta-InverseReinforcementLearning
- CalibratedModel-BasedDeepReinforcementLearning
- Information-TheoreticConsiderationsinBatchReinforcementLearning
- TamingMAML:Controlvariatesforunbiasedmeta-reinforcementlearninggradientestimation
- OptionDiscoveryforSolvingSparseRewardReinforcementLearningProblems
优化类文章
- FingerprintPolicyOptimisationforRobustReinforcementLearning
- CollaborativeEvolutionaryReinforcementLearning
- ComposingValueFunctionsinReinforcementLearning
- Task-AgnosticDynamicsPriorsforDeepReinforcementLearning
- PolicyConsolidationforContinualReinforcementLearning
探索-利用及模型参数
- ExplorationConsciousReinforcementLearningRevisited
- DynamicWeightsinMulti-ObjectiveDeepReinforcementLearning
- ControlRegularizationforReducedVarianceReinforcementLearning
- Dead-endsandSecureExplorationinReinforcementLearning
- Off-PolicyDeepReinforcementLearningwithoutExploration
- Dimension-WiseImportanceSamplingWeightClippingforSample-EfficientReinforcementLearning
- ExtrapolatingBeyondSuboptimalDemonstrationsviaInverseReinforcementLearningfromObservations
- OntheGeneralizationGapinReparameterizableReinforcementLearning
多智能体
- SocialInfluenceasIntrinsicMotivationforMulti-AgentDeepReinforcementLearning
- CURIOUS:IntrinsicallyMotivatedMulti-Task,Multi-GoalReinforcementLearning
- Finite-TimeAnalysisofDistributedTD(0)withLinearFunctionApproximationonMulti-AgentReinforcementLearning
- MaximumEntropy-RegularizedMulti-GoalReinforcementLearning
- Multi-AgentAdversarialInverseReinforcementLearning
- Grid-WiseControlforMulti-AgentReinforcementLearninginVideoGameAI
- QTRAN:LearningtoFactorizewithTransformationforCooperativeMulti-AgentReinforcementLearning
- Actor-Attention-CriticforMulti-AgentReinforcementLearning
图模型强化学习
- TibGM:ATransferableandInformation-BasedGraphicalModelApproachforReinforcementLearning
- SOLAR:DeepStructuredRepresentationsforModel-BasedReinforcementLearning
分布式强化学习
- StatisticsandSamplesinDistributionalReinforcementLearning
- DistributionReinforcementLearningforEfficientExploration
应用类
- ActionRobustReinforcementLearningandApplicationsinContinuousControl
- TransferLearningforRelatedReinforcementLearningTasksviaImage-to-ImageTranslation
- LearningActionRepresentationsforReinforcementLearning
- TheValueFunctionPolytopeinReinforcementLearning
- GenerativeAdversarialUserModelforReinforcementLearningBasedRecommendationSystem
其他
- Kernel-BasedReinforcementLearninginRobustMarkovDecisionProcesses
- ADeepReinforcementLearningPerspectiveonInternetCongestionControl
- ReinforcementLearninginConfigurableContinuousEnvironments
- TighterProblem-DependentRegretBoundsinReinforcementLearningwithoutDomainKnowledgeusingValueFunctionBounds
注：部分文章还没有在arxiv上，或者没有的请自行Google
paper-PDF版本（资源获取）
1、搜索“Deep-RL”或者扫描下面"二维码"，关注本公众号
2、后台回复：ICML2019获取ICML2019深度强化学习相关论文pdf版本。
深度强化学习算法
算法、框架、资料、前沿信息等
长按二维码关注我们吧
上一篇
下一篇
推荐文章
-
自考生课外阅读什么类型的书好
2024-07-02 阅读:770 -
当我们讨论机器学习平台，我们实际上在讨论什么(上)
2019-06-01 阅读:1254 -
汉堡大学：当虚拟筛选遇上机器学习，我们的初心是啥？
2019-05-28 阅读:619 -
偏差、方差、噪声
2019-05-27 阅读:606
自考网版权所有 2024 粤ICP备16065536号