Uncertainty Quantification Metrics for Deep Regression
This work addresses the need for reliable uncertainty quantification in robotics and physical systems, though it is incremental as it focuses on comparing existing metrics rather than introducing new ones.
The paper tackles the problem of evaluating predictive uncertainty metrics for deep regression models in safety-critical applications, finding that Calibration Error is the most stable and interpretable, while recommending against using Spearman's Rank Correlation.
When deploying deep neural networks on robots or other physical systems, the learned model should reliably quantify predictive uncertainty. A reliable uncertainty allows downstream modules to reason about the safety of its actions. In this work, we address metrics for evaluating such an uncertainty. Specifically, we focus on regression tasks, and investigate Area Under Sparsification Error (AUSE), Calibration Error, Spearman's Rank Correlation, and Negative Log-Likelihood (NLL). Using synthetic regression datasets, we look into how those metrics behave under four typical types of uncertainty, their stability regarding the size of the test set, and reveal their strengths and weaknesses. Our results indicate that Calibration Error is the most stable and interpretable metric, but AUSE and NLL also have their respective use cases. We discourage the usage of Spearman's Rank Correlation for evaluating uncertainties and recommend replacing it with AUSE.