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


Current browse context:


Change to browse by:

References & Citations


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

Statistics > Machine Learning

Title: Mixtures of Gaussian Process Experts with SMC$^2$

Abstract: Gaussian processes are a key component of many flexible statistical and machine learning models. However, they exhibit cubic computational complexity and high memory constraints due to the need of inverting and storing a full covariance matrix. To circumvent this, mixtures of Gaussian process experts have been considered where data points are assigned to independent experts, reducing the complexity by allowing inference based on smaller, local covariance matrices. Moreover, mixtures of Gaussian process experts substantially enrich the model's flexibility, allowing for behaviors such as non-stationarity, heteroscedasticity, and discontinuities. In this work, we construct a novel inference approach based on nested sequential Monte Carlo samplers to simultaneously infer both the gating network and Gaussian process expert parameters. This greatly improves inference compared to importance sampling, particularly in settings when a stationary Gaussian process is inappropriate, while still being thoroughly parallelizable.
Subjects: Machine Learning (stat.ML); Machine Learning (cs.LG); Computation (stat.CO)
Cite as: arXiv:2208.12830 [stat.ML]
  (or arXiv:2208.12830v1 [stat.ML] for this version)

Submission history

From: Teemu Härkönen [view email]
[v1] Fri, 26 Aug 2022 18:20:14 GMT (21121kb,D)

Link back to: arXiv, form interface, contact.