<ul><li>Graphical forecasting models learn the structure of time series data via projecting onto a graph.</li><li>Hierarchical Graph Flow (HiGFlow) network introduces a memory buffer variable to store previously seen information across variable resolutions.</li><li>HiGFlow reduces smoothness when mapping onto new feature spaces in the hierarchy.</li><li>Empirical results show that HiGFlow outperforms state-of-the-art baselines, including transformer models, in MAE and RMSE.</li></ul>

Reducing Smoothness with Expressive Memory Enhanced Hierarchical Graph Neural Networks

Discover more