Hi Mike,
if I understand correctly, the FNO only learns from data and does not need the NS-PDE equation. So you need a set of data, in the case of 2D NS, x, y as input variables for the network, and U and V representing the solutions at a given time. The FNO now learns the NS from these data that are subject to these laws. If I see this correctly, you only have to embed the additional variables y (additionally as the input variable for the network) and V into the ‘helper function’. Maybe Conor Daly, the author can made some additional comments on that?
BR, Chris
