INFORMS Open Forum

Khachiyan Prize:  Monique Laurent and Renato Monteiro

Farkas Prize:  Alper Atamturk

Egon Balas Prize:  Alberto del Pia and Stefan Wild

Young Researchers Paper Prize:  Gonzalo Munoz and Felipe Serrano

Student Paper Prize:  Wouter Jongeneel

Many thanks to the prize committees for their diligent work in selecting winners among excellent nominations and submissions from our vibrant optimization community. The prizes will be presented at the Optimization Society's business meeting on Sunday October 15th, 6:30-7:30 pm during the INFORMS 2023 National Meeting, Phoenix, Arizona. Please see the citations and prize committees below.

----------------------------------------------------------------------------

INFORMS Optimization Society 2023 Khachiyan Prize

----------------------------------------------------------------------------

WINNERS:

 Monique Laurent, CWI, Amsterdam, Netherlands

and

 Renato Monteiro, ISyE Georgia Tech, Atlanta, GA.

Citation for Monique Laurent:  For deep structural and algorithmic results in combinatorial, semidefinite, and polynomial optimization, which has led to surprising new mathematical insights at the interface between continuous and discrete optimization, and applications in other areas such as discrete geometry or quantum information.

Citation for Renato Monteiro:  For important theoretical and algorithmic contributions across the range of continuous optimization, including methods for linear and semidefinite programming (especially with low-rank structure), augmented Lagrangian methods, and accelerated proximal-point methods for minimization problems and monotone inclusions, which lead to the optimal high-order schemes.

Prize committee:  Karen Aardal, Michel Goemans (chair), Yurii Nesterov, Stephen Wright.

----------------------------------------------------------------------

INFORMS Optimization Society 2023 Farkas Prize

----------------------------------------------------------------------

WINNER: Alper Atamturk, UC Berkeley, Berkeley, CA.

Citation:  The INFORMS Optimization Society 2023 Farkas Prize is awarded to Alper Atamturk in recognition of his outstanding research in methodology and computer implementation of algorithms for integer programming, nonconvex optimization, and optimization under uncertainty, and for work on applications to network design, logistics, portfolio optimization, and power engineering.    

Alper Atamturk received his PhD in 1998 from Georgia Tech. In the same year, he joined The University of California at Berkeley, where he is now Earl J. Isaac Chair and Chair of the Department of Industrial Engineering and Operations Research (IEOR).

Professor Atamturk's academic work resulted in fundamental results in polyhedral analysis and convexification of problems including mixed-integer conic optimization, submodular optimization, lot sizing, and network design problems. Professor Atamturk's industrial interactions led to the development of highly impactful portfolio management tools. At Berkeley, Professor Atamturk advised many outstanding students and led the development of the IEOR Master of Analytics. A former chair of the INFORMS Optimization Society, Professor Atamturk has served on the editorial boards of several flagship optimization and operations research journals.

Prize committee:  Kurt Anstreicher,  Jesus de Lorea, Simge Kucukyavuz, Nick Sahinidis (chair).

-----------------------------------------------------------------------------

INFORMS Optimization Society 2023 Egon Balas Prize

-----------------------------------------------------------------------------

WINNERS:

 Alberto Del Pia, University of Wisconsin–Madison, Madison, WI

and

 Stefan Wild, Argonne National Laboratory, Lemont, IL.

Citation for Alberto Del Pia:

Alberto Del Pia is an Associate Professor in the Department of Industrial and Systems Engineering and Wisconsin Institute for Discovery at the University of Wisconsin-Madison. He received his PhD from the University of Padova in 2009 and has held postdoctoral positions at Otto-von-Guericke-Universität, ETH Zurich, and as the Herman Golstine Memorial Postdoctoral Fellow at the IBM T.J. Watson Research Center. Dr. Del Pia was awarded the INFORMS Optimization Society Prize for Young Researchers in 2017 and serves on the editorial boards of Mathematical Programming and Discrete Optimization. Dr. Del Pia has made fundamental contributions to a wide range of problems in discrete mathematical optimization. His work has shed light on the power of classes of cutting planes for mixed-integer linear programming, established fundamental theory and algorithms for mixed-integer quadratic programming, and provided deep understanding of the multilinear set which arises in many nonconvex optimization problems. More recently, Dr. Del Pia has been establishing novel connections between discrete optimization theory and problems in machine learning and statistics. The award committee was impressed by the combined depth and breadth of Dr. Del Pia's work.

Citation for Stefan Wild:  

Stefan M. Wild directs the Applied Mathematics and Computational Research Division in the Computing Sciences Area at Lawrence Berkeley National Laboratory. Previously, he was a senior computational mathematician and a deputy division director of the Mathematics and Computer Science Division at Argonne National Laboratory. He received his PhD from Cornell University in 2009. He has served on editorial boards of INFORMS Journal on Computing, Mathematical Programming C, Operations Research, SIAM Journal on Scientific Computing, and SIAM Review. Stefan Wild is internationally known for his work on derivative-free optimization (DFO), from algorithmic development and convergence theory to practical implementations and software tools. He is the co-developer of the so-called data profiles, a technique widely used to measure progress of DFO methods for increasing budgets of function evaluations. He has made key contributions to computational noise estimation and model-based methods for global and non-smooth DFO. Dr. Wild has a long research record of problem solving in optimization involving expensive computer simulations, large data sets, and physical experiments. The award committee was impressed by Dr. Wild's combination of fundamental contributions to optimization theory and algorithms with significant and impactful contributions to scientific computing and applications.

Prize Committee:  Amirali Ahmadi, Dan Bienstock (chair),  Jim Luedtke, Luis Nunes Vicente.

--------------------------------------------------------------------------------------------------

INFORMS Optimization Society 2023 Young Researchers Paper Prize

--------------------------------------------------------------------------------------------------

WINNER:

 Gonzalo Munoz, Universidad de O'Higgins,

and

 Rancagua, Chile and Felipe Serrano, ZIB, Berlin, Germany.

for their joint paper titled "Maximal quadratic-free sets."

Citation: In this paper, Dr. Munoz and Dr. Serrano develop foundations for using cutting plane technology in nonconvex, quadratically constrained quadratic optimization problems. Cutting planes are an indispensable tool in algorithms and solvers for mixed-integer linear optimization. The overall idea is to convexify a nonconvex problem and thus reduce to convex optimization. A central idea from this field is that of intersection cuts. This idea was introduced for mixed-integer linear optimization by Balas in the 70s, but in principle can be applied to a general nonconvex optimization problem with "structured" nonconvexity. Nevertheless, this paper by Dr. Munoz and Dr. Serrano is one of the first ones that applies this technique outside the integer optimization setting, in this case for continuous nonconvex quadratic optimization. There are several highly nontrivial conceptual and technical challenges that must be surmounted to make this transfer of ideas work. First, some convex geometry issues are easier to handle in the integer lattice case because of the regular structure of the integer lattice, and fresh new ideas are needed to handle them in the quadratic case. Second, the structure of quadratics "at infinity" requires developing completely new insights for these tools that have no counterpart in the integer lattice case. The constructions by the authors for handling asymptotes in nonlinear sets also hold promise beyond the quadratic setting. Further, the paper itself is written with a textbook-like exposition, presenting the new ideas with clarity and illustrated with well-designed figures. While the paper itself is of a theoretical and foundational nature, its computational viability has been proven convincingly in follow up work by the authors and other collaborators. In conclusion, we feel this paper makes a fundamental and lasting contribution which has the potential to significantly advance algorithms and solvers for general nonconvex QCQPs, and nonconvex optimization more broadly.

Prize Committee:  Amitabh Basu (chair),  Ilya Hicks, Courtney Paquette, Andy Sun.

---------------------------------------------------------------------------------

INFORMS Optimization Society 2023 Student Paper Prize

---------------------------------------------------------------------------------

WINNER: Wouter Jongeneel,  EPFL, Lausanne, Switzerland

for the paper titled "Small errors in random zeroth-order optimization are imaginary", jointly authored with Man-Chung Yue and Daniel Kuhn.

Citation: The paper proposes a novel approach to zeroth-order optimization for solving continuous minimization problems and opens up new possibilities for numerically stable algorithms. A zeroth-order algorithm is derivative-free as it only makes use of function evaluations. Such methods are needed in simulation-based optimization, minimax optimization, reinforcement learning, etc. and typically belong to one of three classes: direct search methods, model-based methods, and randomized methods. Randomized methods have received much attention due to a recent paper by Nesterov and Spokoiny; the main idea is to choose a direction randomly, then decide how far to move along the direction based on a one-dimensional derivative estimate.  Finite-difference-based derivative estimates suffer from numerical cancellation errors and cannot easily be used to achieve high precision. The approach in the paper by Jongeneel et al. is new and very interesting: It extends the domain of the objective function (assumed to be real-analytic) to a complex domain, then uses ideas from complex analysis to get a derivative estimate with a single function evaluation that is not subject to numerical cancellation. The proposed algorithm is shown to have approximation and convergence guarantees similar to multi-point estimators thereby combining the benefits of multi-point and single-point approaches. Convergence of a zeroth-order algorithm that employs these derivative estimates is proved in convex, strongly convex, and nonconvex settings.  Numerical experiments show that the approach outperforms methods with comparable properties.  The authors show how to apply their method to some problems where function evaluations come from simulations with PDEs. This is impressive. Overall, the paper addresses a fairly general problem and provides an elegant and surprising solution that has the potential to be widely applicable.

Second Place:

 Miaolan Xie (Cornell University, Ithaca, NY) for the paper "Sample Complexity Analysis for Adaptive Optimization Algorithms with Stochastic Oracles" jointly authored with Billy Jin and Katya Scheinberg.

Honorable Mention:

 Aras Selvi (Imperial College London, United Kingdom) for the paper "Differential Privacy via Distributionally Robust Optimization" jointly authored with  Huikang Liu and Wolfram Wiesemann.

Prize committee: Merve Bodur, Frank Curtis, Sanjeeb Dash (chair), Robert Hildebrand, Omar Housni, Giacomo Nannicini, Luis Zuluaga.

------------------------------

Oktay Gunluk

Chair, INFORMS Optimization Society

------------------------------