Open Monthly Meeting – WP5, WP10
These meetings are organized to share and discuss the work that have been done by the WP teams. All partners and network members of TAILOR are welcome to participate!
This seventh meeting will offer two presentations. WP5 (see WP description) scientist Hector Geffner (UPF) will talk about “Top-down representation learning for acting and planning”.
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 without requiring background knowledge, and illustrate the general approach by considering three key learning problems in AI planning: learning general actions models, learning general policies, and learning general subgoal structures (“intrinsic rewards”).
Q&A time should be available after each presentation.
The meeting will end with 20 minutes for general topics, on any aspect of the TAILOR project.
No registration is needed, and the meeting link will be sent by email. (Project partners can find the link here. )
Not sure if you are a member of the TAILOR Network? Not a member yet? More information can be found on the network page.