Donald Goldfarb

The award recognizes the seminal contributions that Donald Goldfarb and Jorge Nocedal have made to the theory and applications of nonlinear optimization over the past several decades. These contributions cover a full range of topics, going from modeling, to mathematical analysis, to breakthroughs in scientific computing. Their work on the variable metric methods (BFGS and L-BFGS, respectively) has been extremely influential.

Goldfarb’s deep research contributions tie together the theoretical and the very practical in traditional linear and nonlinear programming, interior point methods, and the newly in vogue methods developed for signal processing and machine learning; and doing all that through a unique understanding of the fundamental issues in each and all of these areas. His contributions to the field are exceptionally broad, very influential and long-lasting, beginning with the famous Broyden-Fletcher-Goldfarb-Shanno (BFGS) algorithm for nonlinear optimization in the 60’s, then the revolutionary steepest edge simplex method for linear programming in the 80’s and in the last decade first-order methods for large-scale convex optimization. The primal and dual steepest edge simplex algorithms, devised by Goldfarb with Reid and Forrest, respectively, are the most widely used variants of the simplex method. Goldfarb’s work provides the theoretical foundation for many variants of this method implemented in most state-of-the-art commercial linear programming solvers. The Goldfarb-Idnani dual active set method for quadratic programming (QP) is one of the most widely used QP methods.  One should also add that Goldfarb has a knack for exciting both talented PhD students and collaborators, to work on important research problems that he identified.

Nocedal made seminal contributions to the area of unconstrained and constrained nonlinear optimization that have fundamentally reshaped this field. This includes the development of L-BFGS methods, extending interior point methods to non-convex constrained optimization, co-authoring a highly influential book in nonlinear optimization, and recently illuminating the interface between optimization and machine learning via efficient and effective second-order methods. In the 1980s, Nocedal invented the L-BFGS optimization algorithm, the limited memory version of the BFGS method. This opened the door to solving vastly larger unconstrained and box-constrained nonlinear optimization problems than previously possible: Nocedal’s L-BFGS algorithm requires storage that is only a small multiple of the number of variables, whereas the original BFGS method required a quadratic amount of storage. The L-BFGS algorithm has had an immense practical impact, which is difficult to overstate. Nocedal was also instrumental in extending the interior-point revolution beyond convex optimization. In the late 1990s, he and his collaborators proposed the first theoretically sound algorithm for nonlinear and nonconvex optimization problems. This algorithm was practical and, importantly, did not rely on strong assumptions. It should also be added that, all throughout his career, Nocedal has been outstanding at mentoring both students and junior colleagues.

For their fundamental contributions, theoretical and practical, that have, and continue to have, a significant impact on the field of optimization, Donald Goldfarb and Jorge Nocedal are being awarded the John von Neumann Theory Prize of the Institute for Operations Research and the Management Sciences (INFORMS).

Donald Goldfarb has made fundamental contributions to the field of continuous optimization through the design and analysis of innovative algorithms, including the celebrated BFGS quasi-Newton method for nonlinear optimization and the steepest edge simplex method for linear programming.

Alexander Shapiro has been one of the most prolific scholars in the field of Operations Research, contributing significantly to nonlinear analysis (specifically sensitivity and optimality), and to stochastic programming, where his work on complexity analysis and risk-averse decision making has been highly influential.

Selection Committee

Jorge Nocedal (Chair), Michael Todd, Jean-Philippe Vial, Laurence Wolsey.