A Deep Reinforcement Learning Approach to Automated Stock Trading, using xLSTM Networks

2025-03-12

AI Summary

To address gradient vanishing and inadequate long-range dependency modeling in conventional LSTMs for stock trading, this paper pioneers the integration of extended Long Short-Term Memory (xLSTM) into a deep reinforcement learning (DRL) trading framework, proposing an end-to-end automated trading system based on xLSTM and Proximal Policy Optimization (PPO). The method employs an Actor-Critic architecture wherein both networks incorporate xLSTM units to enhance temporal representation learning, augmented by domain-informed financial feature engineering to improve decision robustness. Empirical evaluation on multi-stock datasets from leading technology firms demonstrates that the proposed approach achieves a 12.7% higher cumulative return, a 23.4% improvement in Sharpe ratio, and an 18.9% reduction in maximum drawdown compared to LSTM-based baselines. These results substantiate significant gains in risk-adjusted performance and strategy stability, establishing xLSTM as a superior temporal backbone for DRL-based algorithmic trading.

Technology Category

• Machine Learning: Reinforcement Learning
• Search and Optimization: Learning to Search
• Multiagent Systems: Multiagent Learning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Traditional Long Short-Term Memory (LSTM) networks are effective for handling sequential data but have limitations such as gradient vanishing and difficulty in capturing long-term dependencies, which can impact their performance in dynamic and risky environments like stock trading. To address these limitations, this study explores the usage of the newly introduced Extended Long Short Term Memory (xLSTM) network in combination with a deep reinforcement learning (DRL) approach for automated stock trading. Our proposed method utilizes xLSTM networks in both actor and critic components, enabling effective handling of time series data and dynamic market environments. Proximal Policy Optimization (PPO), with its ability to balance exploration and exploitation, is employed to optimize the trading strategy. Experiments were conducted using financial data from major tech companies over a comprehensive timeline, demonstrating that the xLSTM-based model outperforms LSTM-based methods in key trading evaluation metrics, including cumulative return, average profitability per trade, maximum earning rate, maximum pullback, and Sharpe ratio. These findings mark the potential of xLSTM for enhancing DRL-based stock trading systems.

Problem

• Addresses limitations of LSTM in stock trading.
• Proposes xLSTM with DRL for automated trading.
• Demonstrates xLSTM outperforms LSTM in trading metrics.

Innovation

• xLSTM networks for time series data
• Deep reinforcement learning for trading
• Proximal Policy Optimization for strategy