Semi-supervised Invertible Neural Operators for Bayesian Inverse Problems

Article is Open Access and can be downloaded using the provided link: https://link.springer.com/article/10.1007/s00466-023-02298-8

Keywords:

• Data-driven surrogates
• Invertible neural networks
• Bayesian inverse problems
• Semi-supervised learning

Content:

The scripts provided can be used to generate the results in the paper for the antiderivative case as well as the Reaction-diffusion case.

Dependencies

• Jax 0.4.1
• Pytorch 1.13.1 (only for Dataset generation)

Google Colab (05/2024)

Link to a Google Colab version for the antiderivative test case: https://colab.research.google.com/drive/1t2eTwcTaWX5Jn92VR3ny4eIRBVFS2nqE?usp=sharing

Citation

If this code is relevant for your research, please consider citing:

@article{kaltenbach2023invertibleNeuralOperators,
  title={Semi-supervised Invertible Neural Operators for Bayesian Inverse Problems},
  author={Kaltenbach, Sebastian and Perdikaris, Paris and Koutsourelakis, Phaedon-Stelios},
  journal={Computational Mechanics},
  year={2023},
  publisher={Springer},
  doi = "https://doi.org/10.1007/s00466-023-02298-8",
  url = "https://link.springer.com/article/10.1007/s00466-023-02298-8"
}