n-MeRCI: A new Metric to Evaluate the Correlation Between Predictive Uncertainty and True Error
This addresses the need for reliable uncertainty assessment in robotics applications, where actions depend on confidence, but it is incremental as it builds on existing uncertainty methods.
The paper tackles the problem of evaluating predictive uncertainty in deep learning, particularly for robotics, by proposing a new metric called n-MeRCI to assess the correlation between uncertainty estimates and true error, validated on a toy dataset and monocular depth estimation.
As deep learning applications are becoming more and more pervasive in robotics, the question of evaluating the reliability of inferences becomes a central question in the robotics community. This domain, known as predictive uncertainty, has come under the scrutiny of research groups developing Bayesian approaches adapted to deep learning such as Monte Carlo Dropout. Unfortunately, for the time being, the real goal of predictive uncertainty has been swept under the rug. Indeed, these approaches are solely evaluated in terms of raw performance of the network prediction, while the quality of their estimated uncertainty is not assessed. Evaluating such uncertainty prediction quality is especially important in robotics, as actions shall depend on the confidence in perceived information. In this context, the main contribution of this article is to propose a novel metric that is adapted to the evaluation of relative uncertainty assessment and directly applicable to regression with deep neural networks. To experimentally validate this metric, we evaluate it on a toy dataset and then apply it to the task of monocular depth estimation.