Variational Inference for SDEs Driven by Fractional Noise

paper project page

We present the first variational inference framework for non-Markovian neural stochastic differential equations (SDEs) driven by fractional Brownian Motion (fBM). Our method builds upon the idea of approximating the fBM by a linear combination of Markov processes, driven by the same, Brownian motion. We then provide the variational prior and posterior, as well as the ELBO.