Scaling Dual-Domain Prediction

Background: Effective spatiotemporal prediction models must accurately capture both long-range global trends and abrupt local, high-frequency details to prevent prediction blurring and error accumulation.

Question / Future Work: Further investigation is warranted to explore the application of the proposed wavelet-based, dual-domain gating framework (WaveSFNet) to even higher-resolution or more complex spatiotemporal tasks, particularly those involving extreme non-stationarity or rapid, localized chaotic dynamics where frequency preservation is paramount.