We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

stat.ML

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Statistics > Machine Learning

Title: Deep State-Space Gaussian Processes

Abstract: This paper is concerned with a state-space approach to deep Gaussian process (DGP) regression. We construct the DGP by hierarchically putting transformed Gaussian process (GP) priors on the length scales and magnitudes of the next level of Gaussian processes in the hierarchy. The idea of the state-space approach is to represent the DGP as a non-linear hierarchical system of linear stochastic differential equations (SDEs), where each SDE corresponds to a conditional GP. The DGP regression problem then becomes a state estimation problem, and we can estimate the state efficiently with sequential methods by using the Markov property of the state-space DGP. The computational complexity scales linearly with respect to the number of measurements. Based on this, we formulate state-space MAP as well as Bayesian filtering and smoothing solutions to the DGP regression problem. We demonstrate the performance of the proposed models and methods on synthetic non-stationary signals and apply the state-space DGP to detection of the gravitational waves from LIGO measurements.
Comments: Accepted for publication in Statistics and Computing. The companion code can be found at this https URL
Subjects: Machine Learning (stat.ML); Machine Learning (cs.LG); Methodology (stat.ME)
Journal reference: Statistics and Computing, 2021
DOI: 10.1007/s11222-021-10050-6
Cite as: arXiv:2008.04733 [stat.ML]
  (or arXiv:2008.04733v3 [stat.ML] for this version)

Submission history

From: Zheng Zhao [view email]
[v1] Tue, 11 Aug 2020 14:50:07 GMT (5867kb,D)
[v2] Tue, 24 Aug 2021 07:31:17 GMT (6877kb,D)
[v3] Mon, 6 Sep 2021 13:49:15 GMT (6873kb,D)

Link back to: arXiv, form interface, contact.