2021 Beale — Orchard-Hays Prize Citation
Alberto Costa and Giacomo Nannicini
"RBFOpt: an open-source library for black-box optimization with costly function evaluations"
Mathematical Programming Computation 10 (2018) 597-629.
"On the implementation of a global optimization method for mixed-variable problems"
Open Journal of Mathematical Optimization 2 (2021).
The work described in these papers develops multiple innovations in optimization of expensive derivative-free and black-box functions. The resulting package RBFOpt is notable for allowing continuous, discrete, and categorical variables; automatically learning the best surrogate model class for the objective; speeding convergence by use of low-cost approximations together with full objective evaluations; asynchronously evaluating objective values at different points in parallel; and combining effective global search approaches with multiple local search strategies. These and other ideas are tightly integrated in a computational package that has shown exceptional performance on a variety of problems from different domains. The papers offer an exemplary presentation of this work, combining lucid mathematical and algorithmic development, computational tests showing the benefits of individual innovations, and benchmark comparisons demonstrating the superior performance of RBFOpt as a whole.
Iain Dunning, Joey Huchette, and Miles Lubin
New ideas introduced in this paper touch upon all significant aspects of algebraic modeling language design, and have been successfully integrated into JuMP, a new package for building optimization applications that has attracted users from many fields. Notable innovations of JuMP include efficient embedding of objective and constraint expressions into the Julia programming language through the use of syntactic macros; integration of user-defined functions into model formulations and derivative computations; and a design that facilitates extensions to structured problem classes. The paper is particularly notable for its detailed discussion of the language design and software engineering considerations that had to be addressed in order to combine new ideas effectively into a general-purpose system.
"JuMP: A Modeling Language for Mathematical Optimization"
SIAM Review 59 (2017) 295-320.