Manifold GCN: Diffusion-based Convolutional Neural Network for Manifold-valued Graphs
Researchers have developed new graph neural network layers designed for data residing on Riemannian manifolds. These layers, named Manifold GCN, are based on a diffusion equation and a tangent multilayer perceptron, offering equivariance to node permutations and feature manifold isometries. Initial applications on synthetic data and a real-world Alzheimer's classification task using triangle meshes of the right hippocampus show that these layers outperform existing state-of-the-art networks while being applicable to a broader range of problems. AI
IMPACT Introduces novel graph neural network layers with potential for broader applications in complex data structures.