citiususc/voila: Variational Inference for Langevin Equations

Non-parametric estimation of Langevin equations (also called Stochastic Differential Equations or SDE) from a densely-observed time series. estimate the drift and diffusion terms by modelling them as gaussian processes (GPs). To cope with the computational complexity that calculating the posterior distribution of the GPs requires, the GPs are approximated using a small set of function points, the inducing variables. These inducing variables are the result of evaluating the drift and diffusion terms at some strategically located pseudo-inputs. The pseudo-inputs and the approximate posterior distributions are learnt using variational inference.

Getting started

Package details

AuthorConstantino Antonio Garcia Martinez
MaintainerConstantino Antonio Garcia Martinez <>
LicenseGPL (>= 3)
Package repositoryView on GitHub
Installation Install the latest version of this package by entering the following in R:
citiususc/voila documentation built on May 13, 2019, 7:30 p.m.