Neural Network Reasoning Algorithm of Large-Scale Gragh Based on Parallel Computing

With the continuous development of intelligent computing power, neural networks are widely used in all walks of life. Traditional neural networks, such as convolutional neural networks and fully connected networks, as the carrier of method models, have played an important role in solving many structural data problems. However, for unstructured graph data, it has its own diversity and complexity. For large-scale graphs, it is still challenging to quickly calculate graph neural networks for processing graphs due to the limitations of software and hardware tools and resources. For this reason, this paper proposes a method of multi-GPU block parallel computing based on graph segmentation, so as to improve the computing efficiency of message passing graph neural network in dealing with large-scale graph problems.