Robust Reinforcement Learning for Portfolio Management via Competition and Cooperation Strategies
Primary Area: reinforcement learning
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Keywords: Reinforcement learning, Portfolio management, Competition, Cooperation, Robustness
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.
TL;DR: RoRL via Competition and Cooperation Strategies
Abstract: In this study, we propose an intelligent system for portfolio management that applies robust reinforcement learning within a multi-agent framework. The proposed system incorporates both competition and cooperation strategies to enhance decision-making performance and adaptability. By formulating the portfolio management problem as a cooperative multi-agent environment, agents collaborate and jointly strive to achieve a common goal. On the other hand, the inclusion of competition strategies enables agents to dynamically compete for limited resources and advantages in the market. Specifically, the proposed cooperative strategies employ the absolute value of the reward, prioritizing accelerated model convergence. Meanwhile, the competitive strategies utilize previous rewards to guide action selection, aiming to seek gains and avoid losses. To assess the performance of our model, we evaluate it on a set of real-world financial data. The results obtained demonstrate that the proposed game strategies outperform traditional reinforcement learning approaches.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 7844
Loading