Hybrids of Reinforcement Learning and Evolutionary Computation in Finance: A Survey

Abstract

Many sequential decision-making problems in finance like trading, portfolio optimisation, etc. have been modelled using reinforcement learning (RL) and evolutionary computation (EC). Recent studies on problems from various domains have shown that EC can be used to improve the performance of RL and vice versa. Over the years, researchers have proposed different ways of hybridising RL and EC for trading and portfolio optimisation. However, there is a lack of a thorough survey in this research area, which lies at the intersection of RL, EC, and finance. This paper surveys hybrid techniques combining EC and RL for financial applications and presents a novel taxonomy. Research gaps have been discovered in existing works and some open problems have been identified for future works. A detailed discussion about different design choices made in the existing literature is also included.