Propagating Variational Model Uncertainty for Bioacoustic Call Label Smoothing

TL;DR

This paper introduces an uncertainty-aware label smoothing technique using Bayesian neural networks to improve wildlife call detection accuracy and calibration without costly sampling, by propagating model uncertainty during training.

Contribution

It proposes a novel method for propagating variational model uncertainty for label smoothing, enhancing predictive performance in bioacoustic classification tasks.

Findings

Improved predictive accuracy in wildlife call detection.

Enhanced model calibration through uncertainty-aware training.

Effective end-to-end uncertainty propagation in a variational ResNet.

Abstract

We focus on using the predictive uncertainty signal calculated by Bayesian neural networks to guide learning in the self-same task the model is being trained on. Not opting for costly Monte Carlo sampling of weights, we propagate the approximate hidden variance in an end-to-end manner, throughout a variational Bayesian adaptation of a ResNet with attention and squeeze-and-excitation blocks, in order to identify data samples that should contribute less into the loss value calculation. We, thus, propose uncertainty-aware, data-specific label smoothing, where the smoothing probability is dependent on this epistemic uncertainty. We show that, through the explicit usage of the epistemic uncertainty in the loss calculation, the variational model is led to improved predictive and calibration performance. This core machine learning methodology is exemplified at wildlife call detection, from audio recordings made via passive acoustic monitoring equipment in the animals' natural habitats, with the future goal of automating large scale annotation in a trustworthy manner.