The aim of the symposium is to get together, on-site, an active group of prominent researchers in the area of representation learning for acting and planning, with the purpose of exchanging ideas in a relaxed and friendly atmosphere. Each researcher will have a chance to present their work, to listen and learn from the work of others, and to engage in discussions and activities.
The topics include but are not limited to learning action models, policies, and other high-level representations for acting and planning, using either combinatorial or deep learning approaches, with an emphasis on structure, generalization, transparency, and understanding. Work on reinforcement learning aimed at learning and exploiting structure for generalization is definitely in the scope of the meeting, like neurosymbolic methods applied in the setting of acting and planning.