Uncertainty Quantification with Lightning

[nilsleh]

We are excited to share the first release of the Lightning-UQ-Box a library built on lightning that implements a collection of methods for Uncertainty Quantification (UQ) in Deep Learning models.

One of the main features is that you only have to bring your pytorch model or can use existing ones from libraries like timm or torchseg and the LightningModule written for each UQ Method will take the necessary steps to equip your model with the chosen UQ Method.

As an open-source effort, we hope that this can bring together different approaches from different applications domains, make it easier for practitioners and researchers to include UQ in their projects, and possibly enhance research ideas. Through a series of tutorials, we also hope to give people an entry point that are new to the area of UQ and demonstrate how the library can be potentially useful for their projects.

We would appreciate any feedback, suggestions or of course contributions from the community :)