Deep Hedging under Rough Volatility

TL;DR

This paper evaluates the Deep Hedging framework's effectiveness in rough volatility models, proposing specialized neural network architectures to handle non-Markovian dynamics and comparing hedging performance to jump diffusion models.

Contribution

It introduces neural network architectures tailored for non-Markovian rough volatility models and analyzes their hedging performance and P&L distributions.

Findings

Deep Hedging performs well under rough volatility models.

Specialized architectures better capture non-Markovian features.

Hedging performance is comparable to jump diffusion models at realistic rebalancing frequencies.

Abstract

We investigate the performance of the Deep Hedging framework under training paths beyond the (finite dimensional) Markovian setup. In particular we analyse the hedging performance of the original architecture under rough volatility models with view to existing theoretical results for those. Furthermore, we suggest parsimonious but suitable network architectures capable of capturing the non-Markoviantity of time-series. Secondly, we analyse the hedging behaviour in these models in terms of P\&L distributions and draw comparisons to jump diffusion models if the the rebalancing frequency is realistically small.