English

Non-Factorised Variational Inference in Dynamical Systems

Machine Learning 2018-12-17 v1 Machine Learning

Abstract

We focus on variational inference in dynamical systems where the discrete time transition function (or evolution rule) is modelled by a Gaussian process. The dominant approach so far has been to use a factorised posterior distribution, decoupling the transition function from the system states. This is not exact in general and can lead to an overconfident posterior over the transition function as well as an overestimation of the intrinsic stochasticity of the system (process noise). We propose a new method that addresses these issues and incurs no additional computational costs.

Keywords

Cite

@article{arxiv.1812.06067,
  title  = {Non-Factorised Variational Inference in Dynamical Systems},
  author = {Alessandro Davide Ialongo and Mark van der Wilk and James Hensman and Carl Edward Rasmussen},
  journal= {arXiv preprint arXiv:1812.06067},
  year   = {2018}
}

Comments

6 pages, 1 figure, 1 table

R2 v1 2026-06-23T06:42:54.721Z