1 paper · 1 filter
Blanka Horvath, Josef Teichmann, Zan Zuric
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…