LTS19: NeurIPS 2019 Workshop on Learning Transferable Skills Vancouver Convention Center Vancouver, Canada, December 13-14, 2019 |
Conference website | http://www.skillsworkshop.ai/ |
Submission link | https://easychair.org/conferences/?conf=lts19 |
Submission deadline | September 15, 2019 |
Introducing LTS19
After spending several decades on the margin of AI, reinforcement learning has recently emerged as a powerful framework for developing intelligent systems that can solve complex tasks in real-world environments. This has had a tremendous impact on a wide range of tasks ranging from playing games such as Go and StarCraft to learning dexterity. However, one attribute of intelligence that still eludes modern learning systems is generalizability. Until very recently, the majority of reinforcement learning research involved training and testing algorithms on the same, sometimes deterministic, environment. This has resulted in algorithms that learn policies that typically perform poorly when deployed in environments that differ, even slightly, from those they were trained on. Even more importantly, the paradigm of task-specific training results in learning systems that scale poorly to a large number of (even interrelated) tasks.
Recently there has been a re-invigorated interest in developing learning systems that can learn transferable skills. This could mean robustness to changing environment dynamics, the ability to quickly adapt to environment and task variations or the ability to learn to perform multiple tasks at once (or any combination thereof). This interest has also resulted in a number of new data sets and challenges (e.g. Obstacle Tower Environment, Animal-AI, CoinRun) and an urgency to standardize the metrics and evaluation protocols to better assess the generalization abilities of novel algorithms. We expect this area to continue to increase in popularity and importance, but this can only happen if we manage to build consensus on which approaches are promising, and, equally important, how to test and evaluate them.
We are excited to organize the NeurIPS 2019 Workshop on Learning Transferable Skills. It will be a full-day workshop that includes a mix of invited speakers, peer-reviewed papers (talks and poster sessions) and a panel discussion
Topics Covered
The workshop welcomes both theoretical and applied research, in addition to novel data sets and evaluation protocols. Whilst the majority of the NeurIPS-related work in this area comes from Deep Reinforcement Learning, we will not limit the workshop to only such approaches. More specifically, we welcome papers that include (but are not limited to):
- Novel algorithms and methods which are able to perform:
- Transfer and meta-learning methods that aim to efficiently generalize from a subset of experienced tasks to unseen tasks which may be drawn from the same or a similar distribution as the earlier tasks.
- Multi-task learning methods which aim to explore how agents can simultaneously learn multiple dissociable skills from a given task distribution.
- Unsupervised skill discovery methods which aim to learn diverse behaviors without a specific reward signal.
- Hierarchical learning methods that enable the composition of subgoals or subtasks with shared structure.
- Novel environments, and associated baselines, that enable the study of transferable skills in RL. This may include procedurally generated environments, or environments specifically designed to test for skills which are not context-dependent.
- Metrics and evaluation protocols that can standardize future research outputs around studying and measuring the development of task-independent skills within agents.
- We are also interested in any approaches involving learning or measuring transferable skills that may come from outside the AI community, especially when there are lessons or methodologies that could help facilitate exploration of the topic within AI.
Submission Guidelines
Submissions of technical papers can be up to 8 pages excluding references and appendices. Short or position papers of 2 to 4 pages are welcome. All papers must be submitted in PDF format, using the NeurIPS 2019 author kit. Papers will be peer-reviewed and selected for oral or poster presentations at the workshop. Attendance is open to all, and at least one author of each accepted submission must be present at the workshop.
Reviewing will be blind to the identities of the authors, meaning that your paper should not contain the names or affiliations of the authors. Any supplementary material should be added as appendix to the main paper (authors can only submit a single PDF file).
Submission deadline: 15th of September 23:59 AOE (Anywhere On Earth)
Committees
Program Committee
- TBD
Organizing committee
- Marwan Mattar
- Arthur Juliani
- Danny Lange
- Matthew Crosby
- Benjamin Beyret
Venue
Vancouver Convention Center, Vancouver CANADA
Contact
Any questions regarding submissions should be emailed to lts19.workshop@gmail.com