Using a multi-strain infectious disease model with physical information neural networks to study the time dependence of SARS-CoV-2 variants of concern
Fig 2
Schematic diagram of VOCs-INN for model.
is used to fit the state variables of the model (1) (represented by the pink shaded area), while
is employed to infer the time-varying parameters (represented by the yellow shaded area). Specifically,
and
are used to infer
and
, respectively.
denotes the automatic differentiation operator, and the LOSS consists of two components,
and
. By minimizing the LOSS, simultaneous data fitting and inference of time-varying parameters can be achieved.