Functional GPU Option Pricing at IFL 2019
25 September 2019
Our paper A Functional Approach to Accelerating Monte Carlo based American Option Pricing, with Wojciech Pawlak and Cosmin Oancea, was accepted for presentation at IFL 2019 in Singapore, held on September 25–27.
The paper expresses a demanding simulation algorithm using high-level parallel functions in Futhark. It combines Monte Carlo simulation with least-squares regression to value options that can be exercised before expiry. GPU benchmarks achieve performance comparable to, and in some cases better than, a CUDA implementation optimised by NVIDIA engineers. The functional formulation also makes the algorithm easier to inspect and change.