HERTA proposes a high-efficiency training algorithm for Unfolded Graph Neural Networks, ensuring fast convergence while preserving interpretability.
