Deep Neural Networks as Point Estimates for Deep Gaussian Processes
Abstract
Neural networks and Gaussian processes are complementary in their strengths and weaknesses. Having a better understanding of their relationship comes with the promise to make each method benefit from the strengths of the other. In this work, we establish an equivalence between the forward passes of neural networks and (deep) sparse Gaussian process models. The theory we develop is based on interpreting activation functions as interdomain inducing features through a rigorous analysis of the interplay between activation functions and kernels. This results in models that can either be seen as neural networks with improved uncertainty prediction or deep Gaussian processes with increased prediction accuracy. These claims are supported by experimental results on regression and classification datasets.
Cite
Text
Dutordoir et al. "Deep Neural Networks as Point Estimates for Deep Gaussian Processes." Neural Information Processing Systems, 2021.Markdown
[Dutordoir et al. "Deep Neural Networks as Point Estimates for Deep Gaussian Processes." Neural Information Processing Systems, 2021.](https://mlanthology.org/neurips/2021/dutordoir2021neurips-deep/)BibTeX
@inproceedings{dutordoir2021neurips-deep,
title = {{Deep Neural Networks as Point Estimates for Deep Gaussian Processes}},
author = {Dutordoir, Vincent and Hensman, James and van der Wilk, Mark and Ek, Carl Henrik and Ghahramani, Zoubin and Durrande, Nicolas},
booktitle = {Neural Information Processing Systems},
year = {2021},
url = {https://mlanthology.org/neurips/2021/dutordoir2021neurips-deep/}
}