Overview

FinRL Contest 2024 Website; Starter Kit

Financial Reinforcement Learning (FinRL) is an interdisciplinary field that applies reinforcement learning to perform financial tasks. FinRL’s ability to adapt to changing market conditions makes it a powerful tool for developing strategies in various tasks, such as:

  • Portfolio management

  • Algorithmic trading

  • Option pricing

The FinRL Contest is a competition that explores and evaluates the capability of machine learning methods in finance. FinRL Contest 2024 introduces two tasks designed to address key challenges in FinRL. It features:

  1. Ensemble Learning. Tackling the challenge of policy instability in FinRL, ensemble learning can reduce the variance and bias associated with individual agents, leading to more reliable performance in volatile market conditions. To overcome the sampling bottleneck and accelerate the training of component agents, we provide a vectorized environment that supports massively parallel simulations.

  2. Reinforcement Learning from Market Feedback (RLMF). Large language models (LLMs) have been used in financial tasks, such as sentiment analysis and generating novel trading signals. However, there is a gap between general-purpose LLMs and financial markets, since LLMs trained on general Internet data may not capture the intrinsic dynamics of financial markets. To align LLMs with financial markets, we propose adapting LLMs using Reinforcement Learning from Market Feedback (RLMF), as a counterpart to Reinforcement Learning from Human Feedback (RLHF). RLMF utilizes feedback from the financial market as reward signals, enabling LLMs to learn from actual financial market behaviors.

We design two tasks to reflect these advancements:

  1. Cryptocurrency trading with ensemble methods

  2. LLM-engineered signals with RLMF

We welcome students, researchers, and engineers who are passionate about finance and machine learning. We encourage the development of:

  • Ensemble strategies

  • Novel trading signals

  • Innovative algorithms

These approaches should adapt to changing market conditions and aim to generate superior returns for investors.