EWRL11 (2013)

The 11th European Workshop on Reinforcement Learning (EWRL 2013)

Dates: August 04 − August 09, 2013
Location: Dagstuhl, Germany
Post-Workshop Proceedings.

EWRL11 will be a Dagstuhl Seminar!
This original format for EWRL will no doubt lead to original ideas and fruitful discussions. Read more on the EWRL Dagstuhl Seminar page.

[dates] [schedule] [presentations] [abstracts] [submissions] [logistics] [Dagstuh Reports] [JMLR] [photos]

Important Dates

Conference: August 04 − August 09, 2012
Submission deadline: May 10, 2013

Organizing Committee

Peter Auer (Montan-Universität Leoben, AT)
Marcus Hutter (Australian National University – Canberra, AU)
Laurent Orseau (AgroParisTech – Paris, FR)

Tentative schedule

This schedule is likely to change since most of it will be organized on the fly, as per Dagstuhl Seminar’s spirit.

Sunday afternoon, August, 4th: Arrival
Monday: Exploration/Exploitation
Tuesday: Function Approximation / Policy Search
Wednesday morning: Universal RL
Wednesday afternoon: Hiking/excursion
Thursday morning: POMDP
Thursday afternoon: Other topics 1
Friday morning: Other topics 2
Friday afternoon: Departure

Download the detailed schedule.

Please note that meal break times are tight.

Monday 4th will start with a self-introduction of the participants. You have roughly 1.5 min to state who you are, what you are interested in, and maybe what you will talk about at the workshop. The purpose is to make yourself and your interests/expertise known to the group so that people can approach you more easily.
Note that there are more than 40 participants, which means that 3min/participant would already be too much time.
In this respect, if you intend to use slides, please send us in advance a PDF of no more than 3 slides so that we can compile all slides into one file (and don’t need to switch files between introductions).

The list of participants can be found here.

Tutorials and invited talks

  • Scott Sanner: Symbolic Dynamic Programming
  • Peter Sunehag: Exploration/exploitation
  • Mohammad Ghavamzadeh and Alessandro Lazaric: Statistical Learning Theory in Reinforcement Learning and Approximate Dynamic Programming
  • Marcus Hutter: Universal RL
  • Joëlle Pineau: POMDP

All tutorials are 45min long + 15min for questions.

Oral presentations

Oral presentations are 15min long, with ample time for questions and discussions.
Talks should avoid conference-like reports on past work, and focus on encouraging discussion, raising new questions and interesting ideas.

In addition to presentations for accepted abstracts, we also welcome proposals for oral presentations (10min) for impromptu talks. However, we cannot ensure everyone will have a slot. Participants are encouraged to upload their talk here.

Accepted abstracts

Hierarchical Learning of Motor Skills with Information-Theoretic Policy SearchGerhard Neumann, Christian Daniel, Andras Kupsic, Marc Deisenroth and Jan PetersMulti-objective Reinforcement LearningManuela Ruiz-MontielLearning Skill Templates for Parameterized TasksJan Hendrik Metzen and Alexander FabischRationality in General Reinforcement LearningPeter Sunehag and Marcus HutterABC reinforcement learningChristos Dimitrakakis and Nikolaos TziortziotisSPSA based Actor-Critic Algorithm for Risk Sensitive ControlPrashanth L. A. and Mohammad GhavamzadehTemporal Abstraction in Reinforcement Learning with Proximity StatisticsRico Jonschkowski and Marc ToussaintTowards Active Crowdsourcing for Smart CitiesFrançois Schnitzler, Jia Yuan Yu and Shie MannorRecent Advances in Symbolic Dynamic Programming for Hybrid MDPs and POMDPsScott Sanner and Zahra ZamaniTheoretical Analysis of Planning with OptionsTimothy Mann and Shie MannorPolicy Iteration on Continuous Domains using Rapidly-exploring Random TreesS S Manimaran and Balaraman RavindranPreference-based Evolutionary Direct Policy SearchRobert Busa-Fekete, Balazs Szorenyi, Paul Weng, Weiwei Cheng and Eyke HullermeierSequentially Interacting Markov Chain Monte Carlo Based Policy IterationOrhan Sönmez and A. Taylan CemgilReinforcement Learning using Kernel-Based Stochastic FactorizationJoëlle PineauMethods for Bellman Error Basis Function constructionDoina PrecupReinforcement Learning with Heterogeneous Policy RepresentationsPetar Kormushev and Darwin CaldwellSample Complexity of Multi-task Reinforcement LearningEmma Brunskill and Lihong LiMulti-Objective Reinforcement LearningAnn Nowé, Kristof van Moffaert, Madalina M. DruganSome thoughts on Transfer Learning in RL: on States and RepresentationLutz FrommbergerSolving Simulator-Defined MDPs for Natural Resource ManagementThomas Dietterich, Majid Alkaee Taleghan, Mark Crowley and Sean McGregorRelations Between Reinforcement Learning, Visual Input, Perception and ActionMartijn van OtterloOptimal Universal Explorative AgentsLaurent Orseau, Tor Lattimore and Marcus HutterUniversal RL: Applications and ApproximationsJoel VenessBayesian Reinforcement Learning + ExplorationTor Lattimore and Marcus Hutter

Paper Submission

We are calling for extended abstracts from the entire RL spectrum, may it be for a tutorial, an overview of a topic, the presentation of a novel idea or of recent results or for raising interesting problems or important RL points to be discussed, for either a talk or a poster.

Abstracts must be 2 pages long and must be formatted according to the CoRR formatting guidelines and the general arXiv formatting guidelines. If you are using LaTeX, simply use the default header:
\documentclass[a4paper,12pt]{article}
The PDF is to be submitted through EasyChair (and not to the arXiv or CoRR websites) starting from now and before May 10, 2013.

Important note: Dagstuhl seminars are on invitation only. This means that if there is no space and the paper is not accepted, then unfortunately participation is not possible.

Logistics

Important information was provided to you in an email by “service@dagstuhl.de” on June 21 (GMail link). In particular, it gives you a password that will be referred to in the following by PWD.

We strongly recommend that you carry a printout of the following pages and the aforementioned email with you when you travel to Dagstuhl:

Some more information can be found on the Seminar’s wiki (login 13321, password PWD).
In particular, you can use this page to share information about your arrival times, in case you would like to share a car.

Arrival and departure information:

Kindly check in with the reception staff when you arrive at Schloss Dagstuhl. Dagstuhl’s reception office, located in the newer facility just left of the main building, is open from 3 p.m. to 7 p.m. on Sundays and holidays and from 8 a.m. to 4 p.m. on other days. If reception is closed when you arrive, please use access code PWD followed by F to open the main door. Then follow our self-service check-in procedure as described on the travel information page. If you plan to arrive on any day other than Sunday, please send an email so Dagstuhl can make the appropriate arrangements.

A buffet dinner is available at the Schloss on Sundays from 6 p.m. until the following morning.

Departure is on Friday, August 09, 2013. Dagstuhl requests that you pay your bill for accommodation, meals and your private expenses on the day of your departure before lunch. Dagstuhl accepts cash (euro), MaestroCard, VISA and MasterCard.

Dagstuhl’s postal address:
Schloss Dagstuhl – Leibniz-Zentrum für Informatik
Oktavie-Allee, 66687 Wadern, Germany
phone: +49-6871-9050, fax: +49-6871-905133
mail: lzi@dagstuhl.de

Dagstuhl Reports

Schloss Dagstuhl wants to collect half-page abstracts for each talk we gave, whether invited, impromptu, or based on an accepted extended abstract. This helps them guarantee the quality of their institution, and thus have funding for future seminars. Note that accepted extended abstracts do need to be shortened to half a page (in case this is not possible, please contact the “laurent orseau agroparistech fr”).

To submit your abstract, follow these instructions (the upload page is here but please read the instructions first). You can use the latex template.

The deadline is September 27, 2013.

Note that Dagstuhl Reports is not related to the EWRL11-2013 proceedings, which will probably be with JMLR.

JMLR

JMLR has kindly accepted to host a Special Topic for EWRL-2013.
All participants can submit a paper related to the workshop.

Please follow the author guidelines closely. In particular, you will need to include a cover letter along with your paper, without suggestions for action editors.

When you submit your paper, make sure to select “EWRL” in the drop-down list box at the bottom left corner of the page.

The deadline is December 20th, 2013.

Photos

ERWL11-Group-Photo2Schloss-Dagstuhl1Schloss-Dagstuhl2

Advertisements

%d bloggers like this: