spectral-learning

Proposes a spectral learning method for stochastic nonlinear dynamical systems using deep feature spaces and an operator-based latent state-space model, demonstrating stable performance in forecasting and filtering tasks.

This paper introduces HMH, a hierarchical multi-scale Graph Neural Network framework designed to address oversmoothing and oversquashing in heterophilous graphs. It utilizes spectral filters with Haar bases to achieve scalable learning and improved performance on node and graph classification tasks.