A Simple Probabilistic Model for Uncertainty Estimation

TL;DR

This paper introduces a probabilistic approach for uncertainty estimation in atrial fibrillation detection from ECG signals, improving detection of atypical recordings and confidence in predictions.

Contribution

It proposes predicting beta distribution parameters over class probabilities to enhance uncertainty estimation in ECG classification tasks.

Findings

Improved detection of atypical recordings.

Enhanced confidence calibration of predictions.

Significant accuracy improvements on confident predictions.

Abstract

The article focuses on determining the predictive uncertainty of a model on the example of atrial fibrillation detection problem by a single-lead ECG signal. To this end, the model predicts parameters of the beta distribution over class probabilities instead of these probabilities themselves. It was shown that the described approach allows to detect atypical recordings and significantly improve the quality of the algorithm on confident predictions.