Uncertainty Estimation for Multi-view Data: The Power of Seeing the Whole Picture
It addresses the problem of making neural networks more trustworthy for real-world applications by improving uncertainty estimation in multi-view scenarios, which is an incremental advance over existing unimodal methods.
The paper tackles uncertainty estimation for multi-view data by proposing a new classification framework that combines uncertainty-aware classifiers for each view, resulting in improved accuracy, calibration error, robustness to noise, and out-of-domain detection compared to baselines.
Uncertainty estimation is essential to make neural networks trustworthy in real-world applications. Extensive research efforts have been made to quantify and reduce predictive uncertainty. However, most existing works are designed for unimodal data, whereas multi-view uncertainty estimation has not been sufficiently investigated. Therefore, we propose a new multi-view classification framework for better uncertainty estimation and out-of-domain sample detection, where we associate each view with an uncertainty-aware classifier and combine the predictions of all the views in a principled way. The experimental results with real-world datasets demonstrate that our proposed approach is an accurate, reliable, and well-calibrated classifier, which predominantly outperforms the multi-view baselines tested in terms of expected calibration error, robustness to noise, and accuracy for the in-domain sample classification and the out-of-domain sample detection tasks.