Understanding model uncertainty is important for many applications. We propose Bootstrap Your Own Variance (BYOV), combining Bootstrap Your Own Latent (BYOL), a negative-free Self-Supervised Learning (SSL) algorithm, with Bayes by Backprop (BBB), a Bayesian method for estimating model posteriors. We find that the learned predictive std of BYOV vs. a supervised BBB model is well captured by a Gaussian distribution, providing preliminary evidence that the learned parameter posterior is useful for label free uncertainty estimation. BYOV improves upon the deterministic BYOL baseline (+2.83% test ECE, +1.03% test Brier) and presents better calibration and reliability when tested with various augmentations (eg: +2.4% test ECE, +1.2% test Brier for Salt & Pepper noise).
Cite
@article{arxiv.2312.03213,
title = {Bootstrap Your Own Variance},
author = {Polina Turishcheva and Jason Ramapuram and Sinead Williamson and Dan Busbridge and Eeshan Dhekane and Russ Webb},
journal= {arXiv preprint arXiv:2312.03213},
year = {2023}
}