We congratulate Lindon Roberts who was recently awarded the 2019 Christopher Reddick Prize for his doctoral research “Derivative-free algorithms for nonlinear optimisation problems”.
Awarded by the Centre for Doctoral Training in Industrially Focussed Mathematical Modelling (InFoMM), part of the Mathematical Institute, University of Oxford, the prize recognises the achievements of a doctoral student with ”high quality mathematical outputs, broad engagement with industry, and major involvement in the InFoMM culture”.
Lindon explains derivative-free optimisation (DFO) as an area of computational mathematics concerned with finding extrema of functions when evaluating their derivatives are impractical or inaccurate (which means standard algorithms are not suitable).
“This is particularly relevant in situations where we are trying to optimise a function which is noisy and/or expensive to compute, which arises in areas such as finance, climate science and engineering design” says Lindon.
Through his research, Lindon developed a more efficient DFO method for solving least-squares problems and introduced new algorithmic techniques and software packages which improve the flexibility and robustness of existing DFO methods. He also developed a new DFO method for large-scale least-squares problems.
“This is a rapidly growing research area which has lots of potential benefit for industry. My research was motivated by making DFO techniques more efficient, and to make them more useful in industry”.
Lindon’s research and focus on solving industry challenges, was nurtured during his years at InFoMM. “Through the InFoMM program I was fortunate to have the opportunity to explore an innovative blend of new research methods whilst learning to apply them in real-world settings”.
The project was supported by the Numerical Algorithms Group (NAG), a UK-based not-for-profit technology company. NAG has implemented Lindon’s methods in the NAG Library, a proprietary collection of mathematical software.
In 2019, shortly after completing his doctoral research, Lindon was appointed as an MSI Fellow. Since joining the ANU he has continued his research in DFO, looking further into methods for large-scale problems, as well as applications in image analysis and data science.
Lindon’s doctoral research was supported by academic supervisor Coralia Cartis, University of Oxford and industrial supervisors Jan Fiala and Benjamin Marteau, NAG.