Kees van Kuilenburg

  School of Business and Economics | Bachelor Econometrics and Operations Research

Replicating Portfolios Using Convex Neural Networks

Kees's elevator pitch
Imagine trying to value or hedge a complex financial product, like one with non-linear payouts. Traditionally, this involves intricate calculations that can be time-consuming and error-prone. My thesis explores a novel solution: using machine learning and mathematics to create simplified portfolios of basic financial instruments, like call options, that can replicate these complex payouts. By combining neural networks and optimization techniques, I developed an algorithm that breaks the problem into smaller, manageable parts. This not only makes the process more accurate but also faster and more reliable for real-world financial applications. The approach could significantly enhance tools for traders, risk managers, and financial engineers, making financial markets more efficient and robust.