state-aware-network

DynG-Diff proposes a state-aware dynamic guidance diffusion framework for probabilistic multivariate time series forecasting, improving robustness by adaptively handling variable heterogeneity.