A brief note on understanding neural networks as Gaussian processes
(2021)- Abstract
- As a generalization of the work in [Lee et al., 2017], this note briefly discusses when the prior of a neural network output follows a Gaussian process, and how a neural-network-induced Gaussian process is formulated. The posterior mean functions of such a Gaussian process regression lie in the reproducing kernel Hilbert space defined by the neural-network-induced kernel. In the case of two-layer neural networks, the induced Gaussian processes provide an interpretation of the reproducing kernel Hilbert spaces whose union forms a Barron space.
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/record/7dc2740e-1df7-493a-abe9-3a9a5f3041a8
- author
- Guo, Mengwu LU
- publishing date
- 2021
- type
- Working paper/Preprint
- publication status
- published
- subject
- publisher
- arXiv.org
- language
- English
- LU publication?
- no
- id
- 7dc2740e-1df7-493a-abe9-3a9a5f3041a8
- alternative location
- https://arxiv.org/abs/2107.11892
- date added to LUP
- 2024-03-23 22:08:21
- date last changed
- 2024-04-17 14:45:13
@misc{7dc2740e-1df7-493a-abe9-3a9a5f3041a8, abstract = {{As a generalization of the work in [Lee et al., 2017], this note briefly discusses when the prior of a neural network output follows a Gaussian process, and how a neural-network-induced Gaussian process is formulated. The posterior mean functions of such a Gaussian process regression lie in the reproducing kernel Hilbert space defined by the neural-network-induced kernel. In the case of two-layer neural networks, the induced Gaussian processes provide an interpretation of the reproducing kernel Hilbert spaces whose union forms a Barron space.}}, author = {{Guo, Mengwu}}, language = {{eng}}, note = {{Preprint}}, publisher = {{arXiv.org}}, title = {{A brief note on understanding neural networks as Gaussian processes}}, url = {{https://arxiv.org/abs/2107.11892}}, year = {{2021}}, }