Filters








58,958 Hits in 7.3 sec

Building Adaptive Dialogue Systems Via Bayes-Adaptive POMDPs

Shaowei Png, Joelle Pineau, Brahim Chaib-Draa
2012 IEEE Journal on Selected Topics in Signal Processing  
The main contribution of this paper is to present a Bayesian reinforcement learning framework for learning the POMDP parameters online from data, in a decision-theoretic manner.  ...  We discuss various approximations and assumptions which can be leveraged to ensure computational tractability, and apply these techniques to learning observation models for several simulated spoken dialogue  ...  The empirical return when planning with the 0.65 prior and without further learning is 3.86. The empirical return when planning with the 0. 8 prior and without further learning is 0. 19.  ... 
doi:10.1109/jstsp.2012.2229962 fatcat:56dejd6hgzhoflji32jyzgynka

Multiagent Stochastic Planning With Bayesian Policy Recognition

Alessandro Panella
2013 PROCEEDINGS OF THE THIRTIETH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE AND THE TWENTY-EIGHTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE  
These methods are to be embedded in decision making frameworks for autonomous planning in partially observable multiagent systems.  ...  In my thesis work, I propose methodologies for learning the policy of external agents from their observed behavior, in the form of finite state controllers.  ...  My thesis work proposes a novel framework for autonomous planning in multiagent, partially observable domains with weak knowledge about the other agents' structure.  ... 
doi:10.1609/aaai.v27i1.8506 fatcat:2oaemt3wijhvxfsay4bztw6kxe

An open randomized controlled study comparing an online text-based scenario and a serious game by Belgian and Swiss pharmacy students

Jérôme Berger, Noura Bawab, Jeremy De Mooij, Denise Sutter Widmer, Nicolas Szilas, Carine De Vriese, Olivier Bugnon
2018 Currents in Pharmacy Teaching and Learning  
To compare online learning tools, looped, branch serious game (SG) and linear textbased scenario (TBS), among a sample of Belgian and Swiss pharmacy students.  ...  Participation rate, pre-and post-experience Likert scales and students' clinical knowledge were measured.  ...  Pierreluigi Balabeni for his support in the statistical analysis. We also thank Ms. Elsa Sancey from the University of Geneva and Mr.  ... 
doi:10.1016/j.cptl.2017.11.002 pmid:29764629 fatcat:3gxfouvvuzgu3p2fg6mzlsfbh4

Extensions of POMCP Planning and Applications to Robotic and Cyber-physical Systems

Alberto Castellini, Enrico Marchesini, Giulio Mazzi, Federico Bianchi, Maddalena Zuccotto, Alessandro Farinelli
2021 Zenodo  
We present recent results related to extensions and applications of Partially Observable Monte Carlo Planning to real-world problems involving robotic and cyber-physical systems.  ...  Experiments were performed in the context of two projects: "Dipartimenti di Eccellenza 2018-22", Italian Ministry of Education, Universities and Research, and "Safe Place: Sistemi IoT per ambienti di vita  ...  ACKNOWLEDGMENT The research is partially funded by project "Dipartimenti di Eccellenza 2018-2022", Italian Ministry of Education, Universities and Research, and Safe Place -Sistemi IoT per ambienti di  ... 
doi:10.5281/zenodo.6367936 fatcat:3akdb7jdwjhs3akfkllwil6tty

Extensions of POMCP Planning and Applications to Robotic and Cyber-physical Systems

Alberto Castellini, Enrico Marchesini, Giulio Mazzi, Federico Bianchi, Maddalena Zuccotto, Alessandro Farinelli
2021 Zenodo  
We present recent results related to extensions and applications of Partially Observable Monte Carlo Planning to real-world problems involving robotic and cyber-physical systems.  ...  Experiments were performed in the context of two projects: "Dipartimenti di Eccellenza 2018-22", Italian Ministry of Education, Universities and Research, and "Safe Place: Sistemi IoT per ambienti di vita  ...  ACKNOWLEDGMENT The research is partially funded by project "Dipartimenti di Eccellenza 2018-2022", Italian Ministry of Education, Universities and Research, and Safe Place -Sistemi IoT per ambienti di  ... 
doi:10.5281/zenodo.5900549 fatcat:n4guftvkkfevvfkxk5p76mjihu

Bayesian reinforcement learning for POMDP-based dialogue systems

ShaoWei Png, Joelle Pineau
2011 2011 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)  
In this work, we show that by exploiting certain known components of the system, such as knowledge of symmetrical properties, and using an approximate online planning algorithm, we are able to apply Bayesian  ...  This limitation can be addressed through model-based Bayesian reinforcement learning, which offers a rich framework for simultaneous learning and planning.  ...  In most realistic domains, an exact online planning algorithm is not tractable.  ... 
doi:10.1109/icassp.2011.5946754 dblp:conf/icassp/PngP11 fatcat:hst5crcsuvbczfusvus2uin2de

The influence of prior knowledge on the effectiveness of guided experiment design

Siswa A. N. van Riesen, Hannie Gijlers, Anjo A. Anjewierden, Ton de Jong
2019 Interactive Learning Environments  
Inquiry learning is an effective learning approach if learners are properly guided. Its effectiveness depends on learners' prior knowledge, the domain, and their relationship.  ...  Three conditions were compared in terms of learning gains for learners having distinct levels of prior knowledge.  ...  learning in an online environment and prior knowledge is a very delicate matter that should be treated carefully.  ... 
doi:10.1080/10494820.2019.1631193 fatcat:giqeagwj7fgajdsfa2hirlx4j4

Adapting Plans through Communication with Unknown Teammates

Trevor Sarratt
2016 PROCEEDINGS OF THE THIRTIETH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE AND THE TWENTY-EIGHTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE  
The main contribution of this work is the characterization of the interaction between learning, communication, and planning in ad hoc teams.  ...  In partially observable multiagent domains, agents much share information regarding aspects of the environment such that uncertainty is reduced across the team, permitting better coordination.  ...  techniques may not apply to novel conditions in which no prior knowledge is available.  ... 
doi:10.1609/aaai.v30i1.9816 fatcat:6rhzzdh5gzgfzouxzff225q474

Building Upon What Is Already There: The Role of Prior Knowledge, Background Information, and Scaffolding in Inquiry Learning

Christof Wecker, Ard W. Lazonder, Jennifer L. Chiu, Cheryl Ann Madeira, Jim Slotta, Yvonne Mulder, Ton de Jong, Alexander Rachel, Hartmut Wiesner, Peter Reimann
2012 International Conference of the Learning Sciences  
During the iterative cycles of inquiry learning, learners' prior domain knowledge is modified, refined, and further developed, provided that learners act upon self-assessments of their understanding and  ...  Prior knowledge is one of the most important factors for learning.  ...  Only the lesson planning data will be presented in this paper, in relation to teachers' prior knowledge.  ... 
dblp:conf/icls/WeckerLCMSMJRW012 fatcat:o3ow6vr7lnbnrn7tsrbs2tkfqm

Learning State-Variable Relationships in POMCP: A Framework for Mobile Robots

Maddalena Zuccotto, Marco Piccinelli, Alberto Castellini, Enrico Marchesini, Alessandro Farinelli
2022 Frontiers in Robotics and AI  
Specifically, we focus on Partially Observable Monte Carlo Planning (POMCP) and represent the acquired knowledge with a Markov Random Field (MRF).  ...  We address the problem of learning relationships on state variables in Partially Observable Markov Decision Processes (POMDPs) to improve planning performance.  ...  ., and Cassandra, A. R. (1998). Planning and Acting in Partially Observable Stochastic Domains. Artif.  ... 
doi:10.3389/frobt.2022.819107 pmid:35928541 pmcid:PMC9343685 doaj:99cec9a6c2654afbb39705f031e44e5f fatcat:pft7w73wcvgjdnl37qdksabbju

Bayesian reinforcement learning in continuous POMDPs with gaussian processes

Patrick Dallaire, Camille Besse, Stephane Ross, Brahim Chaib-draa
2009 2009 IEEE/RSJ International Conference on Intelligent Robots and Systems  
In this paper, we consider the problem of optimal control in continuous and partially observable environments when the parameters of the model are unknown.  ...  Our results on the blimp problem show that the approach can learn good models of the sensors and actuators in order to maximize long-term rewards.  ...  This paper aims to investigate a model-based BRL framework that can handle domains that are both partially observable and continuous without assuming any parametric form for the transition, observation  ... 
doi:10.1109/iros.2009.5354013 dblp:conf/iros/DallaireBRC09 fatcat:uj4pakh5lrhanmmmzjglvt6dgi

Model-based Bayesian Reinforcement Learning in Factored Markov Decision Process

Bo Wu, Yanpeng Feng, Hongyan Zheng
2014 Journal of Computers  
A point-based online value iteration approach is then used for planning and learning online.  ...  The experimental and simulation results show that the proposed approach can effectively reduce the number of learning parameters, and enable online learning for dynamic systems with thousands of states  ...  BETTLE [9] and PO-BEETLE [10] are the point-based value iteration algorithms adapted to Bayesian reinforcement learning in fully observable domains and in partially observable domains respectively.  ... 
doi:10.4304/jcp.9.4.845-850 fatcat:ek7223zcdvbjxfcdk5bw2g5sii

Planning Beyond the Sensing Horizon Using a Learned Context [article]

Michael Everett and Justin Miller and Jonathan P. How
2020 arXiv   pre-print
The economic inefficiency of collecting accurate prior maps for navigation motivates the use of planning algorithms that operate in unmapped environments.  ...  with a prior map.  ...  RELATED WORK 1) Planning & Exploration: Classical planning algorithms rely on knowledge of the goal coordinates (A*, RRT) and/or a prior map (PRMs, potential fields), which are both unavailable in this  ... 
arXiv:1908.09171v3 fatcat:ox3p3phugfe5xiyltcncn263vq

Replanning in Domains with Partial Information and Sensing Actions

R. I. Brafman, G. Shani
2012 The Journal of Artificial Intelligence Research  
In this paper we adapt this idea to classical, non-stochastic domains with partial information and sensing actions, presenting a new planner: SDR (Sample, Determinize, Replan).  ...  Replanning via determinization is a recent, popular approach for online planning in MDPs.  ...  Acknowledgement: The authors are grateful to Alexander Albore, Hector Geffner, and Son To for their help in understanding and using their systems.  ... 
doi:10.1613/jair.3711 fatcat:pgesdfl67fau3ax4aobb5yhvkm

Model-based Bayesian Reinforcement Learning for Dialogue Management [article]

Pierre Lison
2013 arXiv   pre-print
The results illustrate in particular the benefits of capturing prior domain knowledge with high-level rules.  ...  This parameter distribution is gradually refined as more data is collected and simultaneously used to plan the agent's actions.  ...  his/her prior knowledge into the domain models.  ... 
arXiv:1304.1819v1 fatcat:2j7gywskjvcm3pmok5eaqwjh2a
« Previous Showing results 1 — 15 out of 58,958 results