A General and Adaptive Robust Loss Function

This directory contains JAX reference code for the paper A General and Adaptive Robust Loss Function, Jonathan T. Barron CVPR, 2019

To use this code, include general.py or distribution.py. general.py implements the "general" form of the loss, which assumes you are prepared to set and tune hyperparameters yourself, and distribution.py implements the probability distribution whose negative log-likelihood corresponds to a shifted "general" loss. This negative log-likelihood used with free parameters for alpha and/or scale corresponds to the "adaptive" loss used in the paper (this code release does not provide a wrapper for this adaptive loss).

This code repository is shared with all of Google Research, so it's not very useful for reporting or tracking bugs. If you have any issues using this code, please do not open an issue, and instead just email jonbarron@gmail.com.

If you use this code, please cite it:

@article{BarronCVPR2019,
  Author = {Jonathan T. Barron},
  Title = {A General and Adaptive Robust Loss Function},
  Journal = {CVPR},
  Year = {2019}
}