Site menu:

picture by

OPT2021

We welcome you to participate in the 13th International (Virtual) OPT Workshop on Optimization for Machine Learning, to be held as a part of the NeurIPS 2021 conference. This year we particularly encourage (but not limit) submissions in the area of Beyond Worst-case Complexity.


We are looking forward to an exciting OPT 2021!

The Workshop

Optimization lies at the heart of many machine learning algorithms. This intimate relation of optimization with ML is the key motivation for the OPT workshop series that aims to bringing together reseachers from both communities in order to promote and generate new interactions.

To foster the spirit of innovation and collaboration OPT 2021 will focus the contributed talks on topics in the area of Beyond Worst-case Complexity:

The algorithms that perform best for training deep neural networks in practice are often not the ones that have the best theoretical performance guarantees. Moreover, increasing research seems to indicate an intimate connection between the optimization algorithm and how well it performs on the test data (generalization). This leads to the question: What new paradigms can help us bridge the gap between optimization theory and practice?

Specifically, we aim to discuss the meanings of generalization, average-case complexity, and other directions that look beyond classical worst-case analysis and to formalize what this means for optimization algorithms.

By bringing together experts in both fields (invited speakers), OPT 2021 will foster insightful discussions around these topics and more.

The OPT2021 Organizers.

Registration

For registration and pricing see NeurIPS.cc.

Program

We have released a preliminary schedule: [google doc], [website].

Registered users can find the schedule and the pre-recorded talks on the NeurIPS.cc virtual site.

Important Dates

(*) If you face severe difficulties meeting this deadline, please contact us before the deadline.