Deep Hedging of Derivatives Using Reinforcement Learning

Cao, Jay; Chen, Jacky; Hull, John C.; Poulos, Zissis

doi:10.3905/jfds.2020.1.052

Cited by 48 publications

(42 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To the best of our knowledge, this paper is the first work to implement the RL algorithms with online learning to hedge contingent claims, particularly variable annuities. Contrary to Xu (2020) and Carbonneau (2021), in which both adapted the state-of-the-art DH approach in Bühler et al (2019), this paper is in line with the recent works by Kolm & Ritter (2019) and Cao et al (2021), while extends with actuarial components. We shall outline the differences between the RL and DH approaches throughout sections 3 and 4, as well as Appendices A and B.…”

Section: Introductionsupporting

confidence: 68%

Pseudo-model-free hedging for variable annuities via deep reinforcement learning

Chong

Cui

2023

Ann. actuar. sci.

View full text Add to dashboard Cite

This paper proposes a two-phase deep reinforcement learning approach, for hedging variable annuity contracts with both GMMB and GMDB riders, which can address model miscalibration in Black-Scholes financial and constant force of mortality actuarial market environments. In the training phase, an infant reinforcement learning agent interacts with a pre-designed training environment, collects sequential anchor-hedging reward signals, and gradually learns how to hedge the contracts. As expected, after a sufficient number of training steps, the trained reinforcement learning agent hedges, in the training environment, equally well as the correct Delta while outperforms misspecified Deltas. In the online learning phase, the trained reinforcement learning agent interacts with the market environment in real time, collects single terminal reward signals, and self-revises its hedging strategy. The hedging performance of the further trained reinforcement learning agent is demonstrated via an illustrative example on a rolling basis to reveal the self-revision capability on the hedging strategy by online learning.

show abstract

Section: Introductionsupporting

confidence: 68%

Pseudo-model-free hedging for variable annuities via deep reinforcement learning

Chong

Cui

2023

Ann. actuar. sci.

View full text Add to dashboard Cite

show abstract

“…Tamar et al (2016) showed how both the first and second moments (and possibly higher moments) of the distribution of G t can be updated using more than one Q-function. Cao et al (2021) used this approach and produced results for an objective function involving the mean and standard deviation of G t , where one Q-function approximates the mean of the terminal return and a second Q-function approximates the expected value of the square of the terminal return. However, the approach is less than ideal.…”

Section: The Rl Modelmentioning

confidence: 99%

“…Previous studies concerned with the application of RL to hedging decisions include Halperin (2017), Buehler et al (2019), Kolm and Ritter (2019), and Cao et al (2021). These authors consider how RL can be used to hedge a single call or put option using a position in the underlying asset.…”

Section: Introductionmentioning

confidence: 99%

Gamma and vega hedging using deep distributional reinforcement learning

Cao

Chen²,

Farghadani³

et al. 2023

Front. Artif. Intell.

Self Cite

View full text Add to dashboard Cite

We show how reinforcement learning can be used in conjunction with quantile regression to develop a hedging strategy for a trader responsible for derivatives that arrive stochastically and depend on a single underlying asset. We assume that the trader makes the portfolio delta-neutral at the end of each day by taking a position in the underlying asset. We focus on how trades in options can be used to manage gamma and vega. The option trades are subject to transaction costs. We consider three different objective functions. We reach conclusions on how the optimal hedging strategy depends on the trader's objective function, the level of transaction costs, and the maturity of the options used for hedging. We also investigate the robustness of the hedging strategy to the process assumed for the underlying asset.

show abstract

“…As a remark of a concept in deep hedging, continuing from the proposition that the approach is model-free, we can include market frictions [3]. Considering the applications of deep reinforcement machine learning methods [6] as typical example models with the presence of market frictions, only generator, loss function and trading instruments are additionally needed [9] that depend on no market dynamics. And Shi et al [3] have developed an algorithm that yields optimum even in high dimensions which primarily depends on the number of hedge instruments [5].…”

Section: Related Workmentioning

confidence: 99%

“…As mentioned by Shi et al [3], the FBSDE solver's algorithm has a problem with the time horizon and discrete-time when scaling. Their approach is targeting the utility function, training and learning the optimal trading strategy directly with reinforcement learning algorithms [6].…”

Section: Introductionmentioning

confidence: 99%

Research on the Applications of Neural Network Algorithms in Deep Hedging

Jin

2023

ACE

View full text Add to dashboard Cite

Under market completeness assumptions, hedging a portfolio of derivatives is straightforward. In view of friction, transaction costs, liquidity and other factors, a framework is presented to extend the pricing and hedging with the hedging strategy treated as a neural network. We study the deep hedging model under incomplete market constraints such as frictions, traction cost, permanent impacts on the market and illiquidity. We discuss the limitations of certain models concerning the applications in deep hedging with constraints. After which, we analyse the advantages of different models and their joint models and find that the hedging strategy is close to the Black-Scholes delta hedging strategy. An example is also given when training after designing two hedging models. The Black-Scholes delta hedging is indeed approximated by unsupervised learning.

show abstract

Deep Hedging of Derivatives Using Reinforcement Learning

Cited by 48 publications

References 24 publications

Pseudo-model-free hedging for variable annuities via deep reinforcement learning

Pseudo-model-free hedging for variable annuities via deep reinforcement learning

Gamma and vega hedging using deep distributional reinforcement learning

Research on the Applications of Neural Network Algorithms in Deep Hedging

Contact Info

Product

Resources

About