Multi-Class Gaussian Process Classification Made Conjugate: Efficient Inference via Data Augmentation

By Théo Galy-Fajou, Florian Wenzel, Christian Donner, Manfred Opper. 2019

32 citations recorded by the source corpus.

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.

Subfields: Artificial Intelligence, Control and Systems Engineering

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Richard S. Zemel
2. Ethan Fetaya
3. Han-Jia Ye
4. Gal Chechik
5. De-chuan Zhan
6. Jake Snell
7. Aviv Navon
8. Idan Achituve

In Control and Systems Engineering

1. Gal Chechik
2. Richard S. Zemel
3. Ethan Fetaya
4. Jake Snell
5. Aviv Navon

Citations in the corpus

• GP-Tree: A Gaussian Process Classifier for Few-Shot Incremental Learning (2021, methodology): “We compared GP-Tree against the following baselines that also used the Pólya-Gamma augmentation to get a conditionally conjugate likelihood: (1) Stick Break Multinomial GP (SBM-GP) (Linderman et al., 2015): that used the stick-breaking process to convert a multinomial likelihood to a product of bin”
• Bayesian Few-Shot Classification with One-vs-Each Pólya-Gamma Augmented Gaussian Processes (2020, methodology): “34 Logistic Softmax GP + Cosine (ML) [8] 60.”