Tuesday November 5, 2020 – 12:15-13:45pm – online
Dirichlet policies for reinforced factor portfolio
Dr. Guillaume Coqueret and Dr. Eric André, emlyon business school
Abstract: This article aims at combining factor investing and reinforcement learning (RL). To this purpose, we resort to a particular policy gradient formulation that relies on firms’ characteristics. Using Dirichlet distributions as the driving policy, we derive closed forms for the gradients which allow to implement REINFORCE methods on a large dataset of US equities. We document that the extreme flexibility permitted by RL is a two-edged sword: while some schemes seem to deliver out-of-sample performance, the range of implementation options yields a large scope of possible outcomes. En route, we confirm the efficacy of equally-weighted investment schemes.
Guillaume is an associate professor of Finance & Data Science at emlyon business school. His research interests revolve around applications of numerical methods in various fields (financial economics mostly). Some of his recent work focuses on supervised learning and on sustainable equity investing.
Eric’s research focus on the consequences of aleatory and epistemic uncertainties on financial markets with applications to risk management and decision making.
Eric holds an engineer’s degree from École Centrale Paris. Before obtaining a PhD in Economics from Aix-Marseille University and joining emlyon business school, he has worked twelve years as an interest rates options trader for some European banks.