Графовые нейронные сети: теория и практика

лекция от Сергея Иванова (research scientist at Criteo) для студентов МФТИ в контексте сетевого подхода в нейронауке. 2021 04 29 In this lecture, we discuss the architecture and convolution of traditional convolutional neural networks. Then we extend it to the graph domain. We will understand the characteristics of graph and will define the graph convolution. Finally, we introduce spectral graph convolutional neural networks and discuss how to perform spectral convolution. We then cover the complete spectrum of Graph Convolutional Networks (GCNs), starting with the implementation of Spectral Convolution through Spectral Networks. It provides insights on applicability of the other convolutional definition of Template Matching to graphs, leading to Spatial networks. Various architectures employing the two approaches are detailed out with their corresponding pros & cons, experiments, benchmarks and applications.