Epidemic monitoring is a burning issue in public health management, as demonstrated by the Covid-19 pandemic. Estimating the intensity of the virus spread is key to take informed decisions, such as lockdown and social distancing, and then to evaluate their efficiency.

To that aim, it is necessary to get accurate estimates of the reproduction number of an epidemic, defined as the number of secondary cases generated by one typical contagious individual, which provides a good quantification of the intensity of the spread. The designed tools not only have to provide good quality estimates, but also have to be robust to low quality data, severely corrupted by adminsitrative noise.

The main objective of this internship is to further robustify previously proposed regularized estimators of the reproduction number by enforcing some spatial consistency between the reproduction number measured in close territories (for example for neighboring French departments). To that aim, a recent methodology based on unbalanced optimal transport will be leveraged to design original spatial regularization, suited to the comparison of non-local patterns. The developed algorithms will then be applied to real Covid-19 data to get better insight in the spatio-temporal dynamics of the virus spread during the entire course of the pandemic.

More informations are available at: https://bpascal-fr.github.io/assets/pdfs/internship2425_optimal-transport-Covid_LS2N.pdf and contacting Barbara Pascal (barbara.pascal@cnrs.fr).