A1756
Title: Deep reinforcement learning and portfolio selection
Authors: Lenka Nechvatalova - Charles University (Czech Republic) [presenting]
Jozef Barunik - UTIA AV CR vvi (Czech Republic)
Abstract: The use of reinforcement learning is proposed to form portfolios for investors with asymmetrical and distorted utility functions. These utility functions do not allow finding optimal portfolio weights as an analytical or straightforward optimization solution. Reinforcement learning is a class of machine learning algorithms where an agent with the goal of maximizing a long-term reward is sequentially making decisions while interacting with the environment and learning from their experience. The portfolio formation is demonstrated on a number of theoretical examples using simulations as well as on empirical datasets. The resulting portfolios are compared with portfolios formed using traditional portfolio selection methods.
