In a proud moment for our community, Ph.D. student Anthony Coache’s innovative research has earned him the Financial Mathematics (FM) paper prize. Awarded by the Society for Industrial and Applied Mathematics (SIAM) it is given once every two years and is open to all students whose paper was accepted or invited for presentation at the SIAM FM conference. This achievement stands testament to the academic rigor fostered within the statistics department.
Coache’s award-winning paper, “Conditionally Elicitable Dynamic Risk Measures for Deep Reinforcement Learning,” was a collaboration with Sebastian Jaimungal, his Ph.D. supervisor at the University of Toronto, and Álvaro Cartea, Director of the Oxford-Man Institute of Quantitative Finance (OMI) at the University of Oxford.
The paper introduced an innovative actor-critic algorithm aimed at solving reinforcement learning problems with a time-consistent, risk-aware agent. The research also uniquely leveraged the concept of (conditional) elicitability, enabling efficient estimation of dynamic spectral risk measures with neural networks. Tested within a statistical arbitrage and portfolio allocation context, the algorithm proved effective with both simulated and real data. The paper is currently under review at the SIAM Journal on Financial Mathematics where it received recommendation for publication after minor revisions.
The paper's development took root during Coache’s six-month research visit to the OMI, an experience he reflects upon fondly. "I had a wonderful experience collaborating with hard-working and motivated researchers from both institutions on problems at the intersection of quantitative finance and machine learning," said Coache.
Despite the challenges of pursuing research during the pandemic and wrestling with the 'impostor syndrome' common to many early-career academics, Coache remained motivated. The recognition from SIAM has given him a significant boost. "Being awarded this prize by experts in the field is extremely rewarding and motivates me to pursue new challenges," he said.
Looking forward, Coache, now entering his fifth year in the Ph.D. program, plans to concentrate on further research projects surrounding control problems and dynamic risk measures. He and Prof. Jaimungal are currently developing a reinforcement learning framework that addresses both risk in a time-consistent manner and model uncertainty.
Join us in congratulating Anthony Coache on this outstanding achievement.