Our fourth symposium will be held on Monday and Tuesday, February 1 to 2, 2016, at Magnus Haus Berlin, Kupfergraben 7, 10117 Berlin
We are in the middle of the SPP’s 2nd phase. The objective of this meeting is to discuss our current and long-term research goals and trends, as well as to foster collaborations. Furthermore, we would like to give time for discussions on a higher level about various topics.
PIs, please prepare a short (10 minutes) presentation to summarize the objectives and results of your project. For each project, please also prepare a simple poster (DIN A1 portrait) for the extended poster session on Monday evening, which will allow for more direct interaction and discussion of your research and technical aspects.
Programme
Monday, February 1, 2016
09:30 – 09:45 Reception with Coffee
09:45 – 10:00 Welcome and information
Marc Toussaint, Marion Lange (Stuttgart/Berlin): The Priority Programme and it’s 2nd phase
10:00 – 11:15 Block 1: Robotics
Tamim Asfour (Karlsruhe), Helge Ritter (Bielefeld), Robert Haschke (Bielefeld): Robots Exploring Tools as Extensions to their Body Autonomously (REBA+)
Sven Behnke (Bonn), Wolfram Burgard (Freiburg): Autonomous Active Object Learning Through Robot Manipulation
Oliver Brock (Berlin), Marc Toussaint (Stuttgart): The Physical Exploration Challenge: Robots Learning to Discover, Actuate, and Explore Degrees of Freedom in the World
Discussion. E.g.: Tools, object affordances, sequential manipulation – there should only be one theory behind all this. If ‘Autonomous Learning’ also means ‘generic learning’, do we achieve this with the current approaches? What are the limitations of our approaches?
11:15 – 11:30 Coffee break
11:30 – 12:30 Discussion session instigated by Helge Ritter
Optimality approaches – are they always our best lanterns to insight in robotics?
12:30 – 13:30 Lunch Buffet at Magnus Haus
13:30 – 15:30 Block 2: Reinforcement Learning
Klaus Obermayer (Berlin): Linking metric and symbolic levels in autonomous reinforcement learning
Nihat Ay (Leipzig): An information theoretic approach to autonomous learning of embodied agents
Gerhard Neumann, Jan Peters (Darmstadt): Learning Modular Policies for Robot Motor Skills
Jan Peters (Darmstadt), Joschka Bödecker (Freiburg): Scalable Autonomous Reinforcement Learning – From scratch to less and less structure
Discussion. E.g.: If it ‘scales’ to a thousand dimensions (in a vector space representation), does that imply that it scales to structured worlds? There is stubborn researchers that believe in some form of planning (on learned models/representations) – do people agree? How does this integrate with the above approaches?
15:30 – 16:00 Coffee and cake
16:00 – 17:00 Discussion session instigated by Wolfram Burgard TBA
17:00 – 18:00 Discussion session on the future of our Priority Programme
Is there some kind of follow up, not necessarily an SPP, but maybe?
(SPP organizational, collaborations, lab rotations, the final summer school)
18:00 Dinner Buffet with Poster Session at Café Wilhelm (Kupfergraben 4a)
Tuesday, February 2, 2016
09:15 – 09:30 Reception with Coffee
09:30 – 11:00 Block 3: Perception and interaction
Bernhard Sick (Kassel): Organic Computing Techniques for Run-Time Self-Adaption of Multi-Modal Activity Recongnition Systems
Daniel Cremers, Rudolf Triebel (München): Efficient Active Online Learning for 3D Reconstruction and Scene Understanding
Barbara Hammer (Bielefeld), Niels Pinkwart (Berlin): Learning Dynamic Feedback in Intelligent Tutoring Systems
Discussion. E.g.: Reinforcement Learning, tutoring/optimal teaching, active querying, learning from demonstration are all forms of ‘social interaction’ – might there be a common framework for such ‘cognitive interaction technologies’?
How can computer vision research and robot perception grow back together again?
11:00 – 11:15 Coffee break
11:15 – 12:15 Discussion session instigated by Tamim Asfour On Dualities and Force
12:15 – 13:15 Lunch Buffet at Magnus-Haus
13:15 – 15:15 Block 4: Deep & auto ML
Reinhold Häb-Umbach (Paderborn): Bayesian Learning of a Hierarchical Representation of Language from Raw Speech
Thomas Brox (Freiburg), Frank Hutter (Freiburg), Philipp Henning (Tübingen): Auto-Tune: Structural Optimization of Machine Learning Frameworks for Large Datasets
Lars Schmidt-Thieme (Hildesheim): Hyperparameter Learning Across Problems
Jörg Lücke (Oldenburg): Autonomous and Efficiently Scalable Deep Learning
Discussion. E.g.: So, does deep learning (or auto ML, variants) now solve everything? When some people say ‘deep learning → AI’, what are they missing?
15:15 – 15:45 Coffee and cake
15:45 – 16:45 Discussion session instigated by Oliver Brock Are roboticists the better deep learners?
16:45 – 17:00 Results, agreements and closing remarks
17:00 End of Symposium