Researchers from Argonne’s Mathematics and Computer Science (MCS) division played a key role in the International Symposium on Mathematical Programming (ISMP) in Bordeaux, France, on July 1–6, 2018.
ISMP is the triennial flagship conference in the field of mathematical optimization. Argonne researchers served on the program committee, organized numerous sessions and were speakers or coauthors of more than 20 presentations. Topics included optimal control, Bayesian optimization, algorithmic differentiation, optimization for power systems, software and the role of high-performance computing, optimization under uncertainty, derivative-free optimization and polynomial optimization.
“This triennial conference provides essential opportunities for Argonne researchers to meet with a broad set of colleagues and collaborators who do not regularly attend conferences in the United States,” said Jeff Larson, an assistant computational mathematician in the MCS division. “Such conferences are important for advancing research and improving the reputation of the Argonne optimization group.”
Some of the activities by MCS division researchers are highlighted below.
Providing Leadership Stefan Wild was on the program/organizing committee and was plenary organizer of the session “Model-Based Methods, Sampling Models, and A New Second-Order Model-Based Method” presented by Luis Nunes Vicente.
In addition, several other MCS researchers served as organizers of various sessions:
- Numerically Efficient Methods for Piecewise Algorithmic Differentiation – Torsten Bosse
- High-Performance Computing in Optimization – Kim Kibaek
- Mixed-Integer PDE-Constrained Optimization – Sven Leyffer
- Software for Nonlinear Optimization – Sven Leyffer
- Nonlinear Optimization with Uncertain Constraints – Charlie Vanaret
- Bayesian and Randomized Optimization – Stefan Wild
“This outstanding participation indicates the leadership role that Argonne plays in this important field,” said MCS division director Valerie Taylor.
Involving all career levels
One of the notable features of MCS participation in this conference was that it included people at various career stages – from postdocs to assistant computer scientists to senior scientists. Here are a few examples.
- Postdocs Xavier Alinson (of Argonne’s Energy Systems Division), Mickael Binois, Brian Dandurand, and Matt Menickelly each gave a presentation. Alinson spoke on multi-row intersection cuts; Binois focused on ways of improving Bayesian optimization; Dandurand explored bilevel optimization approaches for power system security and Menickelly spoke on derivative-free robust optimization.
- Kamil Khan, a recent Director’s Postdoctoral Fellow at Argonne and now an assistant professor of chemical engineering at McMaster University, spoke on generalized derivatives for nonsmooth composite functions.
- Kim Kibaek, an assistant computational mathematician, discussed decomposition methods for mixed-integer programming.
- Sri Hari Krishna Narayanan, a computer scientist, gave a talk on the numerical efficiency of structured abs-normal forms.
- Mihai Anitescu, a senior computational mathematician, presented a talk on exponentially convergent receding horizon constrained optimal control.
Students from around the world also presented work that they had performed during visits to Argonne.
- Predoctoral appointee Mirko Hahn of Germany gave a talk on set-valued steepest descent for binary topology and control optimization. Hahn has also worked with MCS staff on mixed-integer PDE-constrained optimization problems.
- Meenarli Sharma of India spoke on her work on inversion of convection-diffusion partial differential equations. Sharma also has worked with MCS researchers developing the MINOTAUR toolkit for solving mixed-integer nonlinear optimization problems.
- Prashant Palkar of India presented a talk on globally convergent simulation-based convex optimization. Palkar has also collaborated with MCS staff to develop a simulation-based framework to evaluate cyberinfrastructure design choices.
- Anna Thuenen of Germany spoke in a session on Nash equilibria about solving multi-leader-follower games. Thuenen has also worked as a research aide in MCS on mathematical modeling and solution of mixed-integer PDE-constrained optimization.
“It’s great to have early-career researchers working together with MCS world-class scientists. We hope that we provide an inspiration to these students as they pursue their research careers in applied mathematics,” Taylor said.
For further information on the conference and the presentations, see the website: http://ismp2018.sciencesconf.org.