Dear all, please find the announcement below about tomorrow's virtual seminar on topics at the intersection of integer programming and machine learning, as part of the Discrete Optimization Talks (DOTs) series. To receive the Zoom link, please join the mailing list at http://subscribe.discreteopt.com (you should receive the link within two hours of joining). You may need to add "lists@mixedinteger.org" to your approved emails list. However, we have received reports that some subscribers have not received our latest announcement, even in their spam folders. Please email us at talks@discreteopt.com if you are such a subscriber, and we will send you the Zoom link directly.
You may also wish to mark your calendars for the next two DOTs, scheduled for Friday, March 29 and Friday, April 26, both at 12pm ET.
Hope to see you tomorrow,
Aleks
------------------------------------------------------------
DOTs: Discrete Optimization Talks
A Virtual Seminar Series
By the Mixed Integer Programming Society
------------------------------------------------------------
We hope you join us for the next DOTs session this Friday, February 23, 2024 at 12:00pm ET (Zoom link via our mailing list). We will have two half-hour talks (details below) followed by an open informal discussion.
Hope to see you there,
Silvia Di Gregorio (Université Sorbonne Paris Nord)
Aleksandr Kazachkov (University of Florida)
Elias Khalil (University of Toronto)
------------------------------------------------------------
Meeting Time: Friday, February 23, 2024, 12–1pm ET, https://dateful.com/eventlink/1752821086
Meeting Link: [join the mailing list via http://subscribe.discreteopt.com]
Meeting ID: [join the mailing list via http://subscribe.discreteopt.com]
Passcode: [join the mailing list via http://subscribe.discreteopt.com]
------------------------------------------------------------
Speaker: Bartolomeo Stellato (Princeton University)
Title: Differentiable Cutting-plane Layers for Mixed-integer Linear Optimization
Abstract: We consider the problem of optimizing a sequence of mixed-integer linear optimization problems (MIPs) with varying parameters. By combining recent advances in cutting plane generation and differentiation through convex optimization problems, we construct a new differentiable architecture to predict the optimal cutting planes from the key parameters of each instance. During the offline phase, our method maximizes the cutting planes' efficiency by evaluating the derivative of the solution of the continuous relaxations with respect to the cut-generating parameters. We show on preliminary computational results that, once trained, our architecture computes solutions with low infeasibility and suboptimality with fast and predictable execution times.
------------------------------------------------------------
Speaker: Karen Aardal (TU Delft)
Title: Machine-learning augmented branch-and-bound for mixed-integer linear optimization
Abstract: Mixed-integer linear optimization solvers use branch and bound as their main component. In recent years, there has been an explosive development in the use of machine learning algorithms for enhancing all main tasks involved in the branch-and-bound algorithm, such as primal heuristics, branching, cutting planes, node selection and solver configuration decisions. In this talk we mention a selection of results in this area together with some relations between integer linear optimization and deep learning. The talk is based on a joint paper with Lara Scavuzzo Montana, Andrea Lodi, and Neil Yorke-Smith.
------------------------------------------------------------
For more information, including links to past talks, visit our website: https://talks.discreteopt.com. If you have any questions, email us at talks@discreteopt.com.
------------------------------
Aleksandr Kazachkov
Assistant Professor, Department of Industrial and Systems Engineering
Assistant Director, Center for Applied Optimization
University of Florida
https://akazachk.github.io------------------------------