Efficient Deep Hedging mechanism based on value function learning [in Japanese]

Kazuki Matoya, Yunzhuo Wang, Masanori Hirano, Kentaro Imajo

The 38th Annual Conference of the Japanese Society for Artificial Intelligence, p. JSAI2024_4M3GS1002, May 31, 2024

Conference

The 38th Annual Conference of the Japanese Society for Artificial Intelligence

Abstract

DeepHedge, using deep learning and price time series simulation for better hedging, is noted for handling real-world market issues like trading fees, not just ideal markets. It's known that training gets tough with standard feedforward neural networks in Deep Hedging, but some settings have efficient structures like the No-Transaction Band Network. Deep Hedging can be seen through reinforcement learning too. Learning hedging strategies with actor-critic reinforcement learning is done, but this can make training harder. This study introduces an algorithm to model value functions well, making neural network learning easier across many problems. It shows that this method outputs better hedging strategies faster than typical networks.

Keywords

Multi Agent; Financial Market; Micro Foundation;

doi

10.11517/pjsai.JSAI2024.0_4M3GS1002

bibtex

@inproceedings{Matoya2024-jsai38,
  title={{Efficient Deep Hedging mechanism based on value function learning [in Japanese]}},
  author={Kazuki Matoya and Yunzhuo Wang and Masanori Hirano and Kentaro Imajo},
  booktitle={The 38th Annual Conference of the Japanese Society for Artificial Intelligence},
  pages={JSAI2024_4M3GS1002},
  doi={10.11517/pjsai.JSAI2024.0_4M3GS1002},
  year={2024}
}