In addition to invited talks, the workshop will include at least five contributed talks (20 minutes each) and several sessions of contributed lightning talks (5 minutes each). Acceptance of contributed talks and lightning talks will be based on a 2-page abstract that will be reviewed in a double-blind process. Submitted abstracts should be anonymous, without explicit author details (violation of author anonymity will result in rejection). Abstract submissions and their reviews will be carried out using the CMT system at https://cmt3.research.microsoft.com/TOPML2022.
Abstract submission deadline: February 17, 2022, 11:59pm (Anywhere on Earth)
Author notification: March 17, 2022
Format of abstracts: Submitted abstracts should be limited to 2 pages of text, equations, tables, and figures plus additional pages for references. Appendices and supplementary material are not allowed beyond the two page abstract. No specific formatting or style is required besides using standard font size (11pt) and margins (1 in).
When writing your abstracts please note the following points:
1. This workshop is focused on overparameterized learning and will focus on contributions that explicitly consider (approximate or) perfect fitting of training data in ML models. Submitted abstracts should explicitly indicate how their contribution relates to the workshop topic.
2. While the main focus of the workshop is on the theory of overparameterized learning, we also welcome empirical studies that characterize numerical behaviors in ways that can inspire new theoretical studies. Authors of a purely empirical contribution should describe in their abstract the expected impact on the theoretical study of overparameterized learning.
Please note that at least one coauthor of each accepted contribution will be expected to attend the virtual workshop to give a contributed talk or a lightning talk.
The list of accepted contributions will be published on the workshop website. Submitted abstracts will not be published.
Submitted abstracts are allowed to overlap with submissions to other conferences and journals. However, it is the authors’ responsibility to make sure that the other conference or journal allows dual submission to the TOPML workshop.