Deep Reinforcement Learning Algorithms for FinancialDecision-Making
146
by Andrei Neagu
Published onMay 27, 2024
Deep Reinforcement Learning Algorithms for FinancialDecision-Making
·
Abstract
This research aims to develop a benchmarking framework for evaluating Deep Reinforcement Learning (DRL) algorithms in computational finance. Previous work often focuses on a single computational finance task and on a single DRL algorithm, limiting an objective comparison across both different computational finance tasks and different algorithms. Our work establishes a framework containing a variety of state-of-the-art DRL algorithms for a wide range of computational finance tasks and market environments, making objective comparisons more accessible to researchers. We address tasks such as option hedging and pricing, portfolio optimization, and optimal execution and use different classes of DRL algorithms such as policy-based, value-based and actor-critic methods. Our current work in progress focuses on policy-based DRL for option hedging with market impact, where transactions (buying and selling) affect market prices, providing a more realistic market environment than previously published works.
Creative Commons Attribution 4.0 International License (CC-BY 4.0)