Using random testing to manage a safe exit from the COVID-19 lockdown

Markus Mueller; Peter Derlet; Christopher Mudry; Gabriel Aeppli

If on top of the heterogeneity of growth rates one adds finite but weak inter-regional couplings K m =n (t) > 0 (mostly between nearest neighbor regions), one may still expect the eigenvectors of K(t) to be rather localized (a phenomenon well known as Anderson localization [11] in the context of waves propagating in strongly disordered media). By this, one means that the eigenvectors have a lot of weight on few regions only, and little weight everywhere else. That such a phenomenon might occur in the growth pattern of real epidemics is suggested by the significant regional differences in growth rates that we have mentioned above. In such a situation it would seem preferable to adapt restrictive measures to localized regions with strong overlap on unstable eigenvectors of K(t), while minimizing their socio-economic impact in other regions with lower k m (t).