*********************************************************************** * * Informatik-Kolloquium / Computer Science Colloquium * *********************************************************************** Presenter: Hector Geffner (ICREA and Universitat Pompeu Fabra, Barcelona, Spain & Linköping University, Sweden) Title: Language-based representation learning for acting and planning Time: Monday 27-6-2022, 14.30 Location: Room 9222, E3 (Informatik-Zentrum, Erweiterungsbau E3, https://www.informatik.rwth-aachen.de/cms/informatik/Fachgruppe/Informatik-Z...) Abstract: Recent breakthroughs in AI have shown the remarkable power of deep learning and deep reinforcement learning. These developments, however, have been tied to specific tasks, and progress in out-of-distribution generalization has been limited. While it is assumed that these limitations can be overcome by incorporating suitable inductive biases in neural nets, this is left vague and informal, and does not provide meaningful guidance. In this talk, I articulate a different learning approach where representations are learned over domain-independent target languages whose structure and semantics yield a meaningful and strongly biased hypothesis space. The learned representations do not emerge from biases in a low level architecture but from a general preference for the simplest hypothesis that explain the data. I illustrate this general idea by considering three concrete learning problems in AI planning: learning general actions models, learning general policies, and learning general subgoal structures ("intrinsic rewards"). In all these cases, learning is formulated and solved as a combinatorial optimization problem although nothing prevents the use of deep learning techniques instead. Indeed, learning representations over domain-independent languages with a known structure and semantics provides an account of what is to be learned, while learning representations with neural nets provides a complementary account of how representations can be learned. The challenge and the opportunity is to bring the two approaches together. Reference: Target languages (vs. inductive biases) for learning to act and plan. Hector Geffner. AAAI 2022. https://arxiv.org/abs/2109.07195 Short Bio: Hector Geffner is an ICREA Research Professor at the Universitat Pompeu Fabra (UPF) in Barcelona, Spain, and a Wallenberg Guest Professor at Linköping University. He grew up in Buenos Aires and obtained a PhD in Computer Science at UCLA in 1989. He then worked at the IBM T.J. Watson Research Center in NY, USA, and at the Universidad Simon Bolivar, in Caracas. Hector teaches courses on logic, AI, and social and technological change, and is currently doing research on representations learning for acting and planning as part of the ERC project RLeap 2020-2025. Es laden ein: die Dozentinnen und Dozenten der Informatik <http://www.vdaalst.com/> prof.dr.ir. Wil van der Aalst ● <http://www.vdaalst.com/> www.vdaalst.com ● @wvdaalst