# A General and Adaptive Robust Loss Function

## TL;DR

This paper introduces a flexible, adaptive robust loss function that generalizes many existing loss functions, enabling automatic robustness adjustment during training to enhance performance in vision and learning tasks.

## Contribution

It proposes a novel loss function that unifies various robust losses and allows automatic robustness adaptation during training, improving performance without manual tuning.

## Key findings

- Enhanced performance in vision tasks like registration and clustering.
- Improved neural network training for generative image synthesis.
- Automatic robustness adjustment benefits in unsupervised depth estimation.

## Abstract

We present a generalization of the Cauchy/Lorentzian, Geman-McClure, Welsch/Leclerc, generalized Charbonnier, Charbonnier/pseudo-Huber/L1-L2, and L2 loss functions. By introducing robustness as a continuous parameter, our loss function allows algorithms built around robust loss minimization to be generalized, which improves performance on basic vision tasks such as registration and clustering. Interpreting our loss as the negative log of a univariate density yields a general probability distribution that includes normal and Cauchy distributions as special cases. This probabilistic interpretation enables the training of neural networks in which the robustness of the loss automatically adapts itself during training, which improves performance on learning-based tasks such as generative image synthesis and unsupervised monocular depth estimation, without requiring any manual parameter tuning.