Quantification of Uncertainties in Neural Networks (QUiNN) is a python library centered around various probabilistic wrappers over PyTorch modules in order to provide uncertainty estimation in Neural Network (NN) predictions.

Build the library

./build.sh 
or 
./setup.py build; setup.py install

Requirements

numpy, scipy, matplotlib, pytorch

Examples

examples/ex_fit.py
examples/ex_fit_2d.py
examples/ex_ufit.py <method> # where method=mcmc, ens or vi.

Authors

Khachik Sargsyan
Javier Murgoitio-Esandi
Oscar Diaz-Ibarra

Acknowledgements

This work is supported by 
- U.S. Department of Energy, Office of Fusion Energy Sciences (OFES) under Field Work Proposal Number 20-023149.
- Laboratory Directed Research and Development (LDRD) program of Sandia National Laboratories.