# Dynamic Datasets and Market Environments for Financial Reinforcement Learning

**Authors:** Xiao-Yang Liu, Ziyi Xia, Hongyang Yang, Jiechao Gao, Daochen Zha, Ming Zhu, Christina Dan Wang, Zhaoran Wang, Jian Guo  
**arXiv:** [2304.13174](https://arxiv.org/abs/2304.13174)  
**Date:** 2023-04-25  
**Summary page:** https://www.summarizepaper.com/en/paper/2304.13174v1/

## Key points

- Authors Liu, Xia, Yang, Gao, Zha, Zhu, Wang, Wang, and Guo discuss challenges of using deep reinforcement learning in financial markets
- Emphasis on dynamic datasets in finance and introduction of FinRL-Meta as a solution
- FinRL-Meta library offers gym-style market environments for training agents
- Provides examples and reproductions of popular research papers
- Enables visualization of results and performance comparison through community competitions
- Jupyter/Python demos organized into a curriculum available to support the financial reinforcement learning community
- Open-source codes for data curation pipeline can be found on GitHub at https://github.com/AI4Finance-Foundation/FinRL-Meta

## Summary

In their paper titled "Dynamic Datasets and Market Environments for Financial Reinforcement Learning," authors Xiao-Yang Liu, Ziyi Xia, Hongyang Yang, Jiechao Gao, Daochen Zha, Ming Zhu, Christina Dan Wang, Zhaoran Wang, and Jian Guo delve into the challenges faced in utilizing deep reinforcement learning in the financial market. They highlight the unique feature of dynamic datasets in this domain and introduce FinRL-Meta as a solution to these challenges. The library offers gym-style market environments for training FinRL agents and provides examples and reproductions of popular research papers. It also enables users to visualize results and compare performance through community competitions. The authors provide Jupyter/Python demos organized into a curriculum and a documentation website to support the growing community interested in financial reinforcement learning. The open-source codes for the data curation pipeline are available on GitHub at https://github.com/AI4Finance-Foundation/FinRL-Meta.