The paper "Predict Then Propagate: Graph Neural Networks Meet Personalized PageRank" by Johannes Gasteiger, Aleksandar Bojchevski, and Stephan Günnemann introduces a novel approach to semi-supervised node classification on graphs. The authors address the limitation of existing methods, which only consider a small neighborhood of nodes for classification, by leveraging the relationship between graph convolutional networks (GCNs) and PageRank. They propose an improved propagation scheme based on personalized PageRank, which allows for a larger neighborhood to be utilized without leading to oversmoothing. This scheme is integrated into a simple model called Personalized Propagation of Neural Predictions (PPNP) and its fast approximation, APPNP. The models are designed to be computationally efficient and outperform state-of-the-art methods in semi-supervised node classification tasks. The paper includes a thorough experimental evaluation, demonstrating the effectiveness of PPNP and APPNP on multiple datasets, and highlights their advantages in terms of accuracy, training time, and parameter efficiency.The paper "Predict Then Propagate: Graph Neural Networks Meet Personalized PageRank" by Johannes Gasteiger, Aleksandar Bojchevski, and Stephan Günnemann introduces a novel approach to semi-supervised node classification on graphs. The authors address the limitation of existing methods, which only consider a small neighborhood of nodes for classification, by leveraging the relationship between graph convolutional networks (GCNs) and PageRank. They propose an improved propagation scheme based on personalized PageRank, which allows for a larger neighborhood to be utilized without leading to oversmoothing. This scheme is integrated into a simple model called Personalized Propagation of Neural Predictions (PPNP) and its fast approximation, APPNP. The models are designed to be computationally efficient and outperform state-of-the-art methods in semi-supervised node classification tasks. The paper includes a thorough experimental evaluation, demonstrating the effectiveness of PPNP and APPNP on multiple datasets, and highlights their advantages in terms of accuracy, training time, and parameter efficiency.