Summer School on Reinforcement Learning

ACAI Summer School on

Reinforcement Learning

Nieuwpoort, Belgium
7 – 14 October 2017
Sponsored by EurAI and Deepmind

Summer School Info

Dates

7 October 2017 till 14 October 2017

Location

Nieuwpoort, Belgium

Organising Committee Members

Theme

In this summer school we want to go beyond the basic introduction to RL and cover the more advanced topics as well, as it is exactly these more innovative subjects which have resulted in the recent successful applications of RL in areas such as robotics and computer games.

Topics

Following topics will be covered:

  • RL & function approximation
  • RL & speed up through demonstrations, shaping and inverse RL
  • RL & exploration and exploitation aspects (including some theoretical results on regret bounds)
  • RL for Multi-agent systems
  • RL & Deep learning 
  • RL & batch and policy search approaches
  • Recent applications of RL
Concept

We opt for a location which offers full board, to create a real team spirit. Participants (who don’t attent the 2 introdutory days) are expected to arrive on Saturday afternoon, October 7th, in the evening we will organize a social get together. On Sunday morning we plan a poster session by the participants, and in the afternoon an excursion. The actual lectures will start on Monday morning. Lecturers have been asked to include demonstrations and hands on exercises, wherever possible. We also explore the possibility of a friendly contest amongst the participants. The school will be finished by Friday evening, but the accommodation remains available till Saturday morning.

Program

Please check for updates! WHERE:  Follow the directions to  “VERGADERZAAL”, its near to the reception, next the elevator take the stairs down.

RL Introductory Days

Friday 06/10
 [D. Roijers ]
08.00 – 10.00   Breakfast
10.00 – 12.00    Introduction to RL (MDPs, Bellman Equations, Value iteration, Policy iteration)
12.00 – 13.30    Lunch
13.30 – 15.30     Introduction to RL (continuation of morning session)
15.30 – 16.00     Coffee break
16.00 – 18.00    Practical session
19.00                  Dinner

Saturday 07/10
[D. Roijers]
08.00 – 10.00    Breakfast
10.00 – 11.00     Continuation practicals  (participants can decide to start earlier, but people will be available for 10am for assistance)
11.00 – 11.30      Coffee break
11.30 -12.30       Model-based RL, SARSA, Q-learning, actor-critic
12.30 – 14.00     Lunch
14.00 – 15.30     Model-based RL, SARSA, Q-learning, actor-critic
15.30 –  16.00    Coffee break
16.00 – 18.00    Practical session
19.00                  Dinner + Drink at De Vismarkt Bierhuis, Albert I laan 120, 8620 Nieuwpoort

Summer School on Reinforcement Learning

Sunday 08/10
08.00 – 9.30      Breakfast
9.30 -12.00        Official opening of school, Pecha-Kucha by participants.(including             coffee break)
12.00 – 3.30       Lunch
14.00 –  evening Social activity 

  • Departure at 2 pm from the reception desk
  • Walk to Westfront in group (alternative take tram to Nieuwpoort Dorp), tour starts a 3pm
  • Dinner at Den Artiest at 6pm, Kinderlaan 20, Nieuwpoort.

Monday 09/10
08.00 – 9.30       Breakfast
9.30 – 10.30        Eligibility traces [D. Precup]
10.30 -11.00        Coffee break
11.00 – 12.00       Eligibility traces [D. Precup]
12.00 – 14.00       Lunch
14.00 – 15.30       Temporal abstraction [D. Precup]
16.00 – 16.30       Coffee break
16.30 – 18.00       Group work
19.00                     Dinner

Tuesday 10/10
08.00 – 9.30       Breakfast
9.30 – 10.30        Forward search/planning methods [L. Busoniu]
10.30 – 11.00       Coffee break
11.00 – 12.00       Forward search/planning methods [L. Busoniu]
12.00 – 13.00       Lunch
13.00 – 14.30       Forward search/planning methods [L. Busoniu]
15.00 – 15.30       Coffee break
15.30 – 18.00       RL in Robotics [J. Peters]
18.00 – 19.00       Group work
19.00                     Dinner

 

Wednesday 11/10
08.00 – 9.00     Breakfast
9.00 – 10.30      Function Approximation 1 [H.van Hasselt]
10.30 – 11.00     Coffee break
11.00 – 12.00     Function Approximation 2 [H.van Hasselt]
12.00 – 14.00    Lunch
14.00 – 16.00    General intro to Deep RL [H.van Hasselt]
16.00 – 16.30    Coffee break
16.30 – 18.00    Group work
19.00                  Dinner

Thursday 12/10
08.00 – 9.00     Breakfast
9.00 – 10.30      Off Policy RL and Distributional RL  [R. Munos]
10.30 – 11.00     Coffee break
11.00 – 12.30     Off Policy RL and Distributional RL  [R. Munos]
12.30 – 14.00    Lunch (Frietkot)
14.00 – 19.00    Social activity
19.00                  Dinner

Friday 13/10
08.00 – 9.00     Breakfast
9.00 – 10.30      Multi-Agent Learning [V. Lesser]
10.30 – 11.00     Coffee break
11.00 – 12.30     Multi-Agent Learning [V. Lesser]
12.30 – 14.00    Lunch
14.00 – 15.30    Group work
16.00                  Official closure school
18.30                  Dinner

Confirmed lecturers
Accommodation

The accommodation provides two bedroom apartments shared by 2 people (i.e. individual rooms). More Info

Individual accomodation is available upon request and subject to availibility.

Nieuwpoort is easy to reach by public transport. The necessary information will be provided the participants.

Registration fee

The registration fee 670 euro. This registration fee includes full board  accomodation form October 7th till October 14th in a 2 bedroom apartement shared by two participants (i.e. individual rooms).

We also provide a very basic introduction for those students with zero background in RL. This will be optional, and organised on Friday 6th and Saturday 7th .The extra cost will be 150 euro to cover the costs of the 2 extra days of full board accommodation.

Application

Applications should be mailed to Lara Mennes, lmennes@vub.ac.be,  and include : short motivation, 1 page cv and recommendation letter of supervisor (in case of students). Please indicate if you are interested in the two basic RL introductory days. Cutoff days: June 20th 2017 and August 31st 2017.