Current browse context:
stat.ML
Change to browse by:
References & Citations
Statistics > Machine Learning
Title: Compositional Modeling of Nonlinear Dynamical Systems with ODE-based Random Features
(Submitted on 10 Jun 2021 (v1), last revised 25 Oct 2021 (this version, v2))
Abstract: Effectively modeling phenomena present in highly nonlinear dynamical systems whilst also accurately quantifying uncertainty is a challenging task, which often requires problem-specific techniques. We present a novel, domain-agnostic approach to tackling this problem, using compositions of physics-informed random features, derived from ordinary differential equations. The architecture of our model leverages recent advances in approximate inference for deep Gaussian processes, such as layer-wise weight-space approximations which allow us to incorporate random Fourier features, and stochastic variational inference for approximate Bayesian inference. We provide evidence that our model is capable of capturing highly nonlinear behaviour in real-world multivariate time series data. In addition, we find that our approach achieves comparable performance to a number of other probabilistic models on benchmark regression tasks.
Submission history
From: Thomas M. McDonald [view email][v1] Thu, 10 Jun 2021 17:55:13 GMT (1085kb,D)
[v2] Mon, 25 Oct 2021 15:56:02 GMT (3820kb,D)
Link back to: arXiv, form interface, contact.