ACML2023 Workshop: New Horizons of Online Learning

News

• Schedule is available (10/26)
• Call for contributed talks is announced (deadline: 10th October).

Overview

The literature on online learning has been extensively studied over decades, and now we have profound foundations such as online convex optimization and bandit algorithms. Recent developments of AI in related fields provide us with new challenging problems such as multi-objective online optimization, online non-convex optimization, distributed online optimization, and so on, which the standard online learning frameworks might only partially capture. In this workshop, we focus on (but not limited to) the following topics.

• Online decision problems that might not be categorized as standard problems of, e.g., online convex optimization or bandit problems.
• Alternative formulations (such as Blackwell games) and reductions between alternative and standard formulations.
• New applications of online decision-making. In particular, we welcome new applications motivating new formulations of online decision-making.
• Other emerging topics in online learning (e.g., adaptive algorithms that work in both stochastic and adversarial environments).

Invited Speakers

• Shinji Ito (NEC/RIKEN AIP)
• Koji Tabata (Hokkaido Univ.)
• Yaxiong Liu (Hokkaido Univ./RIKEN AIP)

Call for Contributed Talks

We call for contributed talks on the above-listed topics. To encourage stimulating discussions, we also welcome talks on recent previous but original work related to the scope of the workshop (i.e., we allow talk proposals based on previously published original work). Instead, we will not publish any proceedings (but may share abstracts only with participants) and will not treat contributed talks as formal publications.

Submission Instruction: Please E-mail to nhol.workshop@gmail.com with subject line ACML23-NHOL-{paper name} with the following contents.

• If the result is openly published (or in arXiv), include the URL.
• Otherwise, attach short abstract or published paper or manuscript (any format is ok, but short abstract is at least 1 page long).
• Indicating attending face-to-face or online (face-to-face attendance is highly recommended)

Schedule

17:35-

13:30-13:40	Opening remark Kohei Hatano
13:40-14:30	Tutorial: Yaxiong Liu "Blackwell game and online learning"
14:30-15:20	Invited Talk1: Shinji Ito "On best-of-both-worlds online learning algorithms"
15:30-16:00	Coffee Break
16:00-16:50	Invited Talk 2: Koji Tabata "Advancements and Applications of Pure Exploration in Bandit Feedback"
16:50-17:05	Contributed Talk 1: Jongyeong Lee "Optimality of Thompson Sampling with Noninformative Priors for Pareto Bandits"
17:05-17:20	Contributed Talk 2: Ryotaro Mitsuboshi "An Improved Metarounding Algorithm via Frank-Wolfe"
17:20-17:35	Contributed Talk 3: Daniel Ebi "Challenges of Online Decision-Making in Energy Systems"
	Discussion & Conclusion

Organizers

• Kohei Hatano (Kyusu Univ./RIKEN AIP)
• Eiji Takimoto (Kyushu Univ.)
• Atsuyoshi Nakamura (Hokkaido Univ.)
• Junya Honda (Kyoto Univ./RIKEN AIP)

Acknowledgement