Cooperative variance estimation and Bayesian neural networks for disentangling aleatoric and epistemic uncertainties

Real-world data contains aleatoric uncertainty - irreducible noise arising from imperfect measurements or from incomplete knowledge about the data generation process. Mean variance estimation networks can learn this type of uncertainty but require ad-hoc regularization strategies to avoid overfitting and are unable to predict epistemic uncertainty. Bayesian neural networks predict epistemic uncertainty but are difficult to train due to the approximate nature of Bayesian inference. This work cooperatively trains a variance network with a Bayesian neural network to disentangle aleatoric and epistemic uncertainties while improving mean estimation.

Recommended citation: Yi, J. & Bessa, M. A. (2026). Cooperative variance estimation and Bayesian neural networks for disentangling aleatoric and epistemic uncertainties. Proceedings of the 43rd International Conference on Machine Learning (ICML).
Download Paper | Download Bibtex