Using neural network for parameterization of nonlinear interactions in wind wave models

A new neural network (NN) application, a cheap yet accurate approximation for the nonlinear wave-wave interaction term in wind-wave numerical models, has been developed. This NN approximation is 5 orders of magnitude faster than exact calculations. It provides twice better accuracy and is only roughly 3 times slower than approximation used before for calculations of this term in wind-wave models.