1,282 Hits in 1.6 sec

Multi-objectivization of reinforcement learning problems by reward shaping

Tim Brys, Anna Harutyunyan, Peter Vrancx, Matthew E. Taylor, Daniel Kudenko, Ann Nowe
2014 2014 International Joint Conference on Neural Networks (IJCNN)  
In this paper we investigate the multi-objectivization of reinforcement learning problems.  ...  Reward shaping is a technique to speed up reinforcement learning by including additional heuristic knowledge in the reward signal.  ...  Then we formulate the multi-objectivization of a reinforcement learning problem by reward shaping in Section IV, and discuss some theoretical properties thereof.  ... 
doi:10.1109/ijcnn.2014.6889732 dblp:conf/ijcnn/BrysHVTKN14 fatcat:5xgyf7xdsjdfdephtcsp2axipa

Special issue on multi-objective reinforcement learning

Madalina Drugan, Marco Wiering, Peter Vamplew, Madhu Chetty
2017 Neurocomputing  
We also wish to thank the editors of Neurocomputing who supervised an independent review process for those papers for which we had a conflict of interest.  ...  Acknowledgements We would like to thank all of the authors who submitted their work for this issue, as well as the reviewers who generously gave their time and expertise during the review process.  ...  On top of the multi-objectivization mechanism, reward shaping is used to incorporate heuristical knowledge. The goal is to learn the Pareto front of optimal policies.  ... 
doi:10.1016/j.neucom.2017.06.020 fatcat:bw6mnryx3fbitm7rnj6wehopsm

Reinforcement Learning on Multiple Correlated Signals

Tim Brys, Ann Nowé
This extended abstract provides a brief overview of my PhD research on multi-objectivization and ensemble techniques in reinforcement learning.  ...  a linear scalarization (implicit multi-objectivization).  ...  One example of a set of techniques that holds much promise is ensemble techniques for reinforcement learning (Wiering and van Hasselt 2008) .  ... 
doi:10.1609/aaai.v28i1.8773 fatcat:z23vcwamnzfz7akupg5iobyday

Combining Multiple Correlated Reward and Shaping Signals by Measuring Confidence

Tim Brys, Ann Nowé, Daniel Kudenko, Matthew Taylor
This class of problems is very relevant in reinforcement learning, as any single-objective reinforcement learning problem can be framed as such a multi-objective problem using multiple reward shaping functions  ...  Multi-objective problems with correlated objectives are a class of problems that deserve specific attention.  ...  Acknowledgments Tim Brys is funded by a Ph.D grant of the Research Foundation-Flanders (FWO). This work was supported in part by NSF IIS-1149917 and NSF IIS-1319412.  ... 
doi:10.1609/aaai.v28i1.8998 fatcat:s3wv6tvppngrzpbcubpkdtik7e

Risk-sensitivity through multi-objective reinforcement learning

Kristof Van Moffaert, Tim Brys, Ann Nowe
2015 2015 IEEE Congress on Evolutionary Computation (CEC)  
Usually in reinforcement learning, the goal of the agent is to maximize the expected return.  ...  Our approach is based on multi-objectivization where a standard single-objective environment is extended with one (or more) additional objectives.  ...  Multi-objectivization was introduced in reinforcement learning to incorporate multiple pieces of heuristic knowledge in order to speed up learning [4] , [5] .  ... 
doi:10.1109/cec.2015.7257098 dblp:conf/cec/MoffaertBN15 fatcat:oc4wb34t6bfenctn7dz4sdpwwm

A Survey of Reinforcement Learning Techniques: Strategies, Recent Development, and Future Directions [article]

Amit Kumar Mondal
2020 arXiv   pre-print
Reinforcement learning is one of the core components in designing an artificial intelligent system emphasizing real-time response.  ...  First, we illustrated the core techniques of reinforcement learning in an easily understandable and comparable way.  ...  [4] proposed a novel method for the multi-objectivization of Markov Decision Problem through the use of multiple reward shaping functions.  ... 
arXiv:2001.06921v2 fatcat:uwqn4jmginf73ouk3zmm45uozy

Reinforcement learning agents providing advice in complex video games

Matthew E. Taylor, Nicholas Carboni, Anestis Fachantidis, Ioannis Vlahavas, Lisa Torrey
2014 Connection science  
Reinforcement learning transfer using a sparse coded inter-task mapping. In LNAI Post-proceedings of the European Workshop on Multi-agent Systems.  ...  Transfer learning via multiple inter-task mappings.  ...  Multi-objectivization of reinforcement learning problems by reward shaping.  ... 
doi:10.1080/09540091.2014.885279 fatcat:ept2qvn4n5aktjv37ty6vk6vwy

Off-Policy Shaping Ensembles in Reinforcement Learning [article]

Anna Harutyunyan and Tim Brys and Peter Vrancx and Ann Nowe
2014 arXiv   pre-print
This opens up new possibilities for sound ensemble techniques in reinforcement learning. In this work we propose learning an ensemble of policies related through potential-based shaping rewards.  ...  Learning happens in real time, and we empirically show the combination policy to outperform the individual policies of the ensemble.  ...  a multi-objectivization formalism demonstrate its usefullness while treating different shapings as correlated objectives [4] .  ... 
arXiv:1405.5358v1 fatcat:ce7lvoz2knemtaairmh5ndfenu

Hierarchical Potential-based Reward Shaping from Task Specifications [article]

Luigi Berducci, Edgar A. Aguilar, Dejan Ničković, Radu Grosu
2022 arXiv   pre-print
The automatic synthesis of autonomous agents' policies through reinforcement learning relies on the definition of a reward signal that simultaneously captures many, possibly conflicting, requirements of  ...  In this work, we introduce a novel, hierarchical, potential-based reward-shaping approach (HPRS) for defining effective rewards for a large family of problems.  ...  multi-objectivization of the task and solves the multi-objective problem by linear scalarization.  ... 
arXiv:2110.02792v2 fatcat:e7wj377xvffcpegfmuqr3izudi

Transfer learning for direct policy search: A reward shaping approach

Stephane Doncieux
2013 2013 IEEE Third Joint International Conference on Development and Learning and Epigenetic Robotics (ICDL)  
In the perspective of life long learning, a robot may face different, but related situations.  ...  Afterwards, the knowledge base is exploited on a target task, with a reward shaping approach: besides its reward on the task, a policy is credited with a reward computed from the knowledge base.  ...  ACKNOWLEDGMENT This work is supported by the ANR CreAdapt project (ANR-12-JS03-0009).  ... 
doi:10.1109/devlrn.2013.6652568 dblp:conf/icdl-epirob/Doncieux13 fatcat:djiyxqhmmfgwrkmrnlmvrccomi

Reward Tweaking: Maximizing the Total Reward While Planning for Short Horizons [article]

Chen Tessler, Shie Mannor
2020 arXiv   pre-print
Traditionally, this parameter was considered part of the MDP; however, as deep reinforcement learning algorithms tend to become unstable when the effective planning horizon is long, recent works refer  ...  In reinforcement learning, the discount factor γ controls the agent's effective planning horizon.  ...  Broader Impact Reward tweaking tackles a fundamental problem in applied reinforcement learning.  ... 
arXiv:2002.03327v2 fatcat:kucukgik7jdoznquxeb7q6gegy

Optimal Control Based on CACM-RL in a Two-Wheeled Inverted Pendulum

Mariano Gómez, Tomás Arribas, Sebastián Sánchez
2012 International Journal of Advanced Robotic Systems  
CACM-RL in this kind of system.  ...  Learning while maintaining the equilibrium is a complex task. It is easy in stable platforms because the system never reaches an unstable state, but in unstable systems it is very difficult.  ...  Reinforcement Learning Reinforcement learning methods only require a scalar reward (or punishment) to learn to map situations (states) in actions [10] .  ... 
doi:10.5772/54658 fatcat:asws5d22gvgjjd2v4qrih5cfpy

Object Affordance Driven Inverse Reinforcement Learning Through Conceptual Abstraction and Advice

Rupam Bhattacharyya, Shyamanta M. Hazarika
2018 Paladyn: Journal of Behavioral Robotics  
Within human Intent Recognition (IR), a popular approach to learning from demonstration is Inverse Reinforcement Learning (IRL).  ...  An architecture for recognizing human intent is presented which consists of an extended Maximum Likelihood Inverse Reinforcement Learning agent.  ...  Acknowledgement: The authors would also like to thank Zubin Bhuyan, University of Massachusetts Lowell, for the discussion regarding IRL.  ... 
doi:10.1515/pjbr-2018-0021 fatcat:m6a2fm5ja5elnp25prgy6mnjp4

Traditional Narratives of Higher Education Cultural Tensions and Ethical Considerations in Adult Learning

Maryann Krikorian
2020 Journal of Critical Thought and Praxis  
A contemplative model of higher education is shared, supporting meaningful forms of adult learning via compassion/self-compassion, active listening, and mindfulness practices.  ...  With the support of holistic human development and learning theories, I advocate for more integrative approaches to higher education.  ...  For many adult learners, the higher education experience may feel like a high-stakes learning environment to prove self-worth by way of external rewards, eliciting undesirable feelings like that of anxiety  ... 
doi:10.31274/jctp.9744 fatcat:mrkvph7gdzbpfpsqz6dcgkttp4

Real Time Demand Response Modelling for Residential Consumers in Smart Grid Considering Renewable Energy with Deep Learning Approach

S. Sofana Reka, V. Prakash, Hassan Haes Alhelou, Pierluigi Siano, M.E.H. Golshan
2021 IEEE Access  
By adopting deep RL (DRL) approach aforesaid problems can be rectified by enabling end-to-end learning capability of deep neural networks [22] .  ...  Due to recent advancement in artificial intelligence, reinforcement learning(RL) gained more attention to provide a solution to the decision-making problem in smart grid [45] .  ...  In 2019 and 2020 he received the award as Highly cited Researcher by ISI Web of Science Group Engineering, Isfahan University of Technology.  ... 
doi:10.1109/access.2021.3071993 fatcat:5opfbqejqzdqneiffklcetwsam
« Previous Showing results 1 — 15 out of 1,282 results