Deep active subspace data

Data and Jupyter notebooks to reproduce the results of:

Edeling, W. (2023). On the deep active-subspace method. SIAM/ASA Journal on Uncertainty Quantification, 11(1), 62-90.

We applied the deep-active subspace method to:

• An HIV model consisting of 7 coupled ordinary differential equations, with 27 uncertain input parameters.

• A COVID19 model with 51 inputs parameters.

See the paper above for more information.

Contents

To reproduce the results of the HIV model, the following Jupyter notebook are present:

• HIV/HIV.ipynb: reproduce the results of the scalar quantities of interest.

• HIV/HIV_vector.ipynb: reproduce the results of the vector-values quantity of interest.

To reproduce the results for the COVID19 model, run

• COVID19/COVID19.ipynb

All required training data is also present in the HIV and COVID19 directories, see the notebooks for a description.

Funding

This research is funded by the European Union Horizon 2020 research and innovation programme under grant agreement #800925 (VECMA project).