Reproduce “The cross-sectional stock return predictions via quantum neural network and tensor network” on OpenResearch

About this paper

Nozomu Kobayashi, Yoshiyuki Suimon, Koichi Miyamoto, Kosuke Mitarai

Researchers applied quantum neural networks (QCL) and quantum-inspired tensor networks (TN) to predict cross-sectional stock returns in the Japanese market. The Tensor Network model achieved the highest annualized excess return of 3.71% and an information ratio of 0.69, surpassing classical linear regression and neural network benchmarks, while QCL demonstrated competitive performance, particularly in recent market conditions.

Key insights

• Quantum-inspired Tensor Network models can achieve superior risk-adjusted returns in stock prediction with fewer parameters than classical neural networks.
• Quantum Circuit Learning, a quantum neural network, demonstrated improved performance in later market environments, suggesting potential robustness against overfitting compared to classical neural networks.
• Both quantum and quantum-inspired models effectively capture non-linear relationships and interaction effects in financial data that are missed by traditional linear models.

Summary by alphaXiv.