Predicting Electromagnetic Field (EMF) Exposure Using Heterogeneous Graph Neural Networks

In this paper, we provide a novel heterogeneous graph neural network (GNN)-based approach to predict the electromagnetic field (EMF) exposure distribution in an indoor environment. The prediction model leverages the concept of heterogeneous graph that allows multiple distinct types of nodes and edges existing to represent a physical indoor environment, providing an improved accuracy compared to homogeneous GNN-based counterpart.