Graph-Informed Neural Networks for Regressions on Graph-Structured Data

In this work, we extend the formulation of spatial-based graph convolutional networks with a new architecture, called graph-informed neural network (GINN). This architecture is specifically designed for regression tasks on graph-structured data that are not suitable well-known networks, such as functions domain and codomain defined two sets values vertices graph. particular, formulate (GI) layer exploits adjacent matrix given to define unit connections in describing convolution operation inputs associated We study GINN models respect maximum-flow test problems stochastic flow networks. GINNs show very good abilities interesting potentialities. Moreover, conclude by real-world application flux problem underground fractures.