Mathematical Finance, Stochastic Analysis, and Machine Learning Seminar by Ivan Gvozdanovic: Time Consistent Reinforcement Learning for Optimal Consumption under Epstein-Zin Preferences

Speaker: Ivan Gvozdanovic, Illinois Institute of Technology 

Title: Time Consistent Reinforcement Learning for Optimal Consumption under Epstein-Zin Preferences

Abstract: In this talk, I present a new class of least squares reinforcement learning algorithms for optimal consumption under elasticity of intertemporal substitution and risk aversion preferences. First, we cast the optimal consumption problem as a discrete time Markov Decision Process, and then derive a least-squares Q-Learning algorithm suitable for non-linear monotone certainty equivalents. In order to benchmark its policy estimation convergence properties, we compare the solution against a Least Squares Monte-Carlo and binomial tree methods. Finally, we demonstrate our least-squares Q-learning algorithm on an optimal consumption problem applied to SPDR S\&P 500 ETF Trust (SPY) data. This is a joint work with Professor Matthew Dixon.

 

Mathematical Finance, Stochastic Analysis, and Machine Learning