The Aleatoric Uncertainty Estimation Using a Separate Formulation with Virtual Residuals
This addresses uncertainty quantification for regression tasks, offering a method to improve accuracy in applications like age and depth estimation, but it appears incremental as it builds on existing uncertainty estimation frameworks.
The paper tackles the problem of aleatoric uncertainty underestimation in regression by proposing a separable formulation that decouples target and uncertainty estimation, and it demonstrates superior performance over a state-of-the-art technique in experiments on simulation data, age estimation, and depth estimation.
We propose a new optimization framework for aleatoric uncertainty estimation in regression problems. Existing methods can quantify the error in the target estimation, but they tend to underestimate it. To obtain the predictive uncertainty inherent in an observation, we propose a new separable formulation for the estimation of a signal and of its uncertainty, avoiding the effect of overfitting. By decoupling target estimation and uncertainty estimation, we also control the balance between signal estimation and uncertainty estimation. We conduct three types of experiments: regression with simulation data, age estimation, and depth estimation. We demonstrate that the proposed method outperforms a state-of-the-art technique for signal and uncertainty estimation.