neural operators extend data-driven models to map between
infinite-dimensional functional spaces. While these operators perform
effectively in either the time or frequency domain, their performance may be
limited when applied to non-stationary spatial or temporal signals whose
frequenc