The 30th meeting of Special Interest Group on Financial Informatics of Japanese Society for Artificial Intelligence, pp. 51-57, Mar. 4, 2023
The 30th meeting of Special Interest Group on Financial Informatics of Japanese Society for Artificial Intelligence (SIG-FIN)
Deep hedging, a framework for hedging a portfolio of derivatives using deep learning and price time-series simulation, has been gaining popularity because it allows for more realistic trading strategies that consider market frictions, such as transaction costs. However, for deep hedging learning, a specific price process, such as the Heston process, is often assumed to generate a fictional underlying asset price. In this study, we propose a new method for enabling deep hedging learning without the specific price process. In the proposed method, adversarial learning between deep hedging and price-series generator is employed. Thanks to the adversarial learning, no assumption for the price process is necessary for deep hedging learning. According to the results of our experiments, we showed that our proposed method could perform hedging almost as well as the original deep hedging.
Deep Hedging; Adversarial Learning; Option; Price Process;
10.11517/jsaisigtwo.2023.FIN-030_51
@inproceedings{Hirano2023-sigfin30,
title={{Adversarial Deep Hedging Free from the Assumption of Underliers' Price Process [in Japanese]}},
author={Masanori HIRANO and Kentaro MINAMI and Kentaro IMAJO},
booktitle={The 30th meeting of Special Interest Group on Financial Informatics of Japanese Society for Artificial Intelligence},
pages={51-57},
doi={10.11517/jsaisigtwo.2023.FIN-030_51},
year={2023}
}