Deep signature FBSDE algorithm

<p style='text-indent:20px;'>We propose a deep signature/log-signature FBSDE algorithm to solve forward-backward stochastic differential equations (FBSDEs) with state and path dependent features. By incorporating the transformation into recurrent neural network (RNN) model, our shortens training time, improves accuracy, extends time horizon comparing methods in existing literature. Moreover, algorithms can be applied wide range of applications such as option pricing involving high-frequency data, model ambiguity, games, which are linked parabolic partial (PDEs), path-dependent PDEs (PPDEs). Lastly, we also derive convergence analysis algorithm.</p>