Constant Time Graph Neural Networks

The recent advancements in graph neural networks (GNNs) have led to state-of-the-art performances various applications, including chemo-informatics, question-answering systems, and recommender systems. However, scaling up these methods huge graphs, such as social Web remains a challenge. In particular, the existing for accelerating GNNs either are not theoretically guaranteed terms of approximation error or incurred at least linear time computation cost. this study, we reveal query complexity uniform node sampling scheme Message Passing Neural Networks, GraphSAGE, attention (GATs), convolutional (GCNs). Surprisingly, our analysis reveals that method is completely independent number nodes, edges, neighbors input depends only on tolerance confidence probability while providing theoretical guarantee error. To best knowledge, first article provide within constant time. Through experiments with synthetic real-world datasets, investigated speed precision validated results.