Deep Hedging under Rough Volatility

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