uncertainty baselines

We have hosted the application uncertainty baselines in order to run this application in our online workstations with Wine or directly.

Quick description about uncertainty baselines:

Uncertainty Baselines is a collection of strong, well-documented training pipelines that make it straightforward to evaluate predictive uncertainty in modern machine learning models. Rather than offering toy scripts, it provides end-to-end recipes�data input, model architectures, training loops, evaluation metrics, and logging�so results are comparable across runs and research groups. The library spans canonical modalities and tasks, from image classification and NLP to tabular problems, with baselines that cover both deterministic and probabilistic approaches. Techniques include deep ensembles, Monte Carlo dropout, temperature scaling, stochastic variational inference, heteroscedastic heads, and out-of-distribution detection workflows. Each baseline emphasizes reproducibility: fixed seeds, standard splits, and strong metrics such as calibration error, AUROC for OOD, and accuracy under shift.

Features:

• End-to-end, reproducible pipelines for uncertainty evaluation
• Coverage of ensembles, MC dropout, SVI, and calibration methods
• Standardized metrics for OOD detection and calibration quality
• Baselines across vision, language, and tabular tasks
• Clear configuration files and logging for fair comparisons
• Strong defaults that can be extended for new research ideas

Programming Language: Python.
Categories: