A Data-Driven Measure of Relative Uncertainty for Misclassification Detection
This addresses the need for better uncertainty estimation in machine learning models to improve reliability, though it appears incremental as it builds on existing misclassification detection methods.
The paper tackles the problem of misclassification detection by introducing a data-driven measure of relative uncertainty that identifies unreliable predictions based on soft-prediction patterns, showing empirical improvements over state-of-the-art methods in image classification tasks.
Misclassification detection is an important problem in machine learning, as it allows for the identification of instances where the model's predictions are unreliable. However, conventional uncertainty measures such as Shannon entropy do not provide an effective way to infer the real uncertainty associated with the model's predictions. In this paper, we introduce a novel data-driven measure of uncertainty relative to an observer for misclassification detection. By learning patterns in the distribution of soft-predictions, our uncertainty measure can identify misclassified samples based on the predicted class probabilities. Interestingly, according to the proposed measure, soft-predictions corresponding to misclassified instances can carry a large amount of uncertainty, even though they may have low Shannon entropy. We demonstrate empirical improvements over multiple image classification tasks, outperforming state-of-the-art misclassification detection methods.