Call for Challenge Questions

We are soliciting “challenge questions” from authors at the workshop and other junior researchers! After each invited speaker’s 25-minute talk, a junior researcher will present a 1-minute contributed challenge question, followed by a response from the invited speaker, and finally the audience question period. The goal is to give junior researchers a bit of visibility and the opportunity to interact with a more senior researcher within the workshop program itself.

If your challenge question is selected, we will ask you to prepare one slide to be appended to the invited speaker’s presentation; after the speaker’s invited talk, you will be called upon to pose your question, and the invited speaker will answer verbally. The audience question period will follow. As references, please see the end of Doina Precup’s, Pieter Abbeel’s and Jan Peters’ invited talks at the Hierarchical Reinforcement Learning Workshop at NeurIPS 2017.

Challenge Question Deadline: Sunday, May 5th, 2019, 12:00 PM noon CDT

Please submit your challenge question via this Google form.

Call for Extended Abstracts

A powerful solution to the problem of generalization and sample complexity in reinforcement learning (RL) is the deliberate use of inductive bias. There has been a recent resurgence of interest in methods of imposing or learning inductive bias in RL in the form of structure and priors, including, for example, prior distributions for Bayesian inference, learned hyperparameters in a multi-task or meta-learning setup, or structural constraints such as temporal abstraction or hierarchy.

The goal of this workshop is to bring together researchers across a variety of domains, including RL and machine learning practitioners, neuroscientists, and cognitive scientists, to discuss the role that structure and priors play in RL. We invite the submission of abstracts on topics including, but not limited to:

  • Bayesian inference as used in RL
  • meta-RL
  • transfer learning in RL
  • modularity and compositionality
  • hierarchical RL
  • temporal abstraction
  • structured state and action abstractions

We also invite abstracts that address the following questions directly:

  • What is the trade-off between generality and the use of structure and priors in RL, in the context of specific tasks or in general, and how can we evaluate this in practice?

  • What are the practical or theoretical implications of specific ways of imposing or learning structure or priors in RL?

  • How can we learn data-driven structure and priors for RL (via transfer in RL, meta-RL, or multi-task RL)?

  • How can the different communities (including cognitive science, neuroscience, and machine learning) benefit from collaborative research on these topics?

Important Deadlines

Extended abstract deadline: Thursday, March 7th, 2019, 11:59 PM anywhere on Earth
Decision notification: Friday, March 29th, 2019, 1:00 PM anywhere on Earth
Camera-ready deadline: Thursday, May 2nd, 2019, 11:59 PM anywhere on Earth

Abstract Format

Extended abstracts should be a short research paper of at most 5 pages long (excluding references) in the LaTeX format defined by this package (adapted from the ICLR 2019 conference format). The main text should include all text and figures, and should adequately describe the work, its contributions, and its limitations. Abstracts may include a supplement of up to 12 pages, but reviewers are not required to read any supplementary material. Abstracts must be anonymized.

Please submit your extended abstract via CMT by the deadline given above.

Selection Criteria

Work that has already appeared or is scheduled to appear in a journal, workshop, or conference (including ICLR 2019) must be significantly extended to be eligible for workshop submission. Work that is currently under review at another venue or has not yet been published in an archival format as of the date of the deadline may be submitted. This includes submissions to the ICML conference proceedings, which are welcome.

All submissions will undergo peer review by the workshop’s program committee. Accepted abstracts will be chosen based on technical merit, empirical validation, novelty, and suitability to the workshop’s goals.

Presentation Details

All accepted abstracts will be presented in the form of a poster. A few select contributions will additionally be given as contributed talks. Accepted papers will be posted in a non-archival format on the workshop website.