Call for Talks


Important updates

NEW!   The list of accepted talks is now available here!

This call is also available in PDF format

CLAI Unconf is a ContinualAI effort to complement existing events and draw inspiration from standard practice in various scientific disciplines. The conference will be an entirely virtual, multi-time-zone, and affordable event with priority on inclusion, diversity, and accessibility. The program will cover interdisciplinary perspectives on Continual Learning through various initiatives. Following disruptive ideas and inspired by common practice in many natural sciences, CLAI Unconf contributions will be based on proposals for talks to highlight current and upcoming research trends. Thus, ideas can be presented and discussed within the community independently of conventional publication mechanisms.

In the spirit of an unconference, the ContinualAI Unconference (CLAI Unconf) is largely community driven, while maintaining the highest standards of scientific communication. To this end, CLAI Unconf invites researchers of all levels (that is, students, junior, and senior researchers) to participate in the Unconf by submitting a proposal for a short or long contributed talk (5 and 15 minutes respectively). Proposals should be in the form of a single-page abstract, and the presenter’s name should be clearly identifiable in the document. There are no strict requirements for the layout of the single-page abstract, however, the font size should be at least equal to or bigger than Computer Modern Roman (cmr) size 10 with single spacing. Submitted abstracts should outline the proposed talk and summarize the content that will be presented. We encourage talks to cover prospective ideas, recent trends, challenges, positions, and visions, promote discussions, or serve educational purposes. We discourage talks to focus specifically on a single paper, but tying together multiple works would be fine.

  • Deadline: 18th of August 2023

Submitted abstracts will receive a light review that will mainly focus on suitability of the proposed topic. If more abstracts will be received than can be accommodated, the review process might also be used to select those abstract that are deemed to be of most interest to the expected audience.

If your submitted abstract is selected, the talk should be delivered live during the Unconference. The selected proposals will be divided in groups based on topics as well as preferred timezone(s).

As some non-exhaustive examples, topics could range from online learning and non-stationary environments to (deep) continual learning or reasoning. They could take a neuroscience perspective, delve into applications such as robotics, vision or language, or focus on theoretical and philosophical advances. Similar to the call for pre-registrations, we welcome papers across many disciplines, including but not limited to more specific topics within:

  • (Deep) learning beyond static datasets: deep continual representation learning and foundation models, distributed learning over time and multi-agent systems, semi-supervised and unsupervised continual learning, meta-learning and continual reinforcement learning; 
  • Online learning and learning in non-stationary environments: adaptation and knowledge accumulation, learning with fixed features, computational and memory efficiency;
  • Discovery of novel elements over time: out-of-distribution detection, drift detection, open world learning, handling partially observable information from potentially disparate sources; 
  • Continual learning in the brain: neuroscientific perspectives, memory consolidation and learning in the brain, new insights from and for cognitive science, bio-inspired methods;
  • Practical continual learning applications: robotics, vision,  language learning, predictive maintenance, AutoML and MLOps, continuum edge-cloud; and
  • Critical analyses & positions: evaluation, experimental protocols, fairness & ethical considerations, terminology consolidation, documentation, complex data collection systems.

Attendance

CLAI Unconf is a fully virtual event with multiple time zones to accommodate participants from all around the world.