Gaussian Process Classification Using Posterior Linearization
(2019) In IEEE Signal Processing Letters 26(5). p.735-739- Abstract
- This letter proposes a new algorithm for Gaussian process classification based on posterior linearization (PL). In PL, a Gaussian approximation to the posterior density is obtained iteratively using the best possible linearization of the conditional mean of the labels and accounting for the linearization error. PL has some theoretical advantages over expectation propagation (EP): all calculated covariance matrices are positive definite and there is a local convergence theorem. In experimental data, PL has better performance than EP with the noisy threshold likelihood and the parallel implementation of the algorithms.
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/record/f4c05d8a-ae81-429d-8443-8819789fcce9
- author
- Garcia-Fernandez, Angel F ; Tronarp, Filip LU and Särkkä, Simo
- publishing date
- 2019
- type
- Contribution to journal
- publication status
- published
- subject
- in
- IEEE Signal Processing Letters
- volume
- 26
- issue
- 5
- pages
- 735 - 739
- publisher
- IEEE - Institute of Electrical and Electronics Engineers Inc.
- external identifiers
-
- scopus:85064572097
- ISSN
- 1070-9908
- DOI
- 10.1109/LSP.2019.2906929
- language
- English
- LU publication?
- no
- id
- f4c05d8a-ae81-429d-8443-8819789fcce9
- date added to LUP
- 2023-08-20 22:46:58
- date last changed
- 2023-11-10 10:56:01
@article{f4c05d8a-ae81-429d-8443-8819789fcce9, abstract = {{This letter proposes a new algorithm for Gaussian process classification based on posterior linearization (PL). In PL, a Gaussian approximation to the posterior density is obtained iteratively using the best possible linearization of the conditional mean of the labels and accounting for the linearization error. PL has some theoretical advantages over expectation propagation (EP): all calculated covariance matrices are positive definite and there is a local convergence theorem. In experimental data, PL has better performance than EP with the noisy threshold likelihood and the parallel implementation of the algorithms.}}, author = {{Garcia-Fernandez, Angel F and Tronarp, Filip and Särkkä, Simo}}, issn = {{1070-9908}}, language = {{eng}}, number = {{5}}, pages = {{735--739}}, publisher = {{IEEE - Institute of Electrical and Electronics Engineers Inc.}}, series = {{IEEE Signal Processing Letters}}, title = {{Gaussian Process Classification Using Posterior Linearization}}, url = {{http://dx.doi.org/10.1109/LSP.2019.2906929}}, doi = {{10.1109/LSP.2019.2906929}}, volume = {{26}}, year = {{2019}}, }