Modeling COVID-19 spread in the USA using metapopulation SIR models coupled with graph convolutional neural networks
The episode discusses a paper modeling COVID-19 spread in the USA using a hybrid approach developed by high school student Petr Kisselev and professor Padmanabhan Seshaiyer. The authors combined the classic SIR model with a graph convolutional neural network, allowing them to predict state-by-state infections more accurately than standard models. This method also enables real-time estimation of the reproduction number (R0).