Multi-Class Gaussian Process Classification Made Conjugate: Efficient Inference via Data Augmentation
Abstract
We propose a new scalable multi-class Gaussian process classification approach building on a novel modified softmax likelihood function. The new likelihood has two benefits: it leads to well-calibrated uncertainty estimates and allows for an efficient latent variable augmentation. The augmented model has the advantage that it is conditionally conjugate leading to a fast variational inference method via block coordinate ascent updates. Previous approaches suffered from a trade-off between uncertainty calibration and speed. Our experiments show that our method leads to well-calibrated uncertainty estimates and competitive predictive performance while being up to two orders faster than the state of the art.
Cite
Text
Galy-Fajou et al. "Multi-Class Gaussian Process Classification Made Conjugate: Efficient Inference via Data Augmentation." Uncertainty in Artificial Intelligence, 2019.Markdown
[Galy-Fajou et al. "Multi-Class Gaussian Process Classification Made Conjugate: Efficient Inference via Data Augmentation." Uncertainty in Artificial Intelligence, 2019.](https://mlanthology.org/uai/2019/galyfajou2019uai-multiclass/)BibTeX
@inproceedings{galyfajou2019uai-multiclass,
title = {{Multi-Class Gaussian Process Classification Made Conjugate: Efficient Inference via Data Augmentation}},
author = {Galy-Fajou, Théo and Wenzel, Florian and Donner, Christian and Opper, Manfred},
booktitle = {Uncertainty in Artificial Intelligence},
year = {2019},
pages = {755-765},
volume = {115},
url = {https://mlanthology.org/uai/2019/galyfajou2019uai-multiclass/}
}