Skip to content
Log in
#graph neural networks Review Open access

Learning on Edges: A Narrative Review of Graph Neural Networks from Recursive Networks to Geometric Deep Learning

Zen Revista 10 IA
Aug 2026 · Zenodo (CERN European Organization for Nuclear Research)
Advanced Graph Neural Networks

Abstract

Graph neural networks---learning over relational, irregular structure by passing messages between nodes---generalized deep learning's grids to the graph: molecules, social networks, knowledge bases, and the web. This article presents a narrative review of that arc's canonical line: Sperduti and Starita's 1997 structure classification, Gori, Monfardini, and Scarselli's 2005 graph-domain learning, Scarselli and colleagues' 2009 GNN model, Bruna and colleagues' 2014 spectral networks, Defferrard and colleagues' 2016 localized filtering, Kipf and Welling's 2017 graph convolutions, Gilmer and colleagues' 2017 message passing, Hamilton, Ying, and Leskovec's 2017 GraphSAGE, Velickovic and colleagues' 2018 attention, Ying and colleagues' 2019 GNNExplainer, Wu and colleagues' 2021 comprehensive survey, and Bronstein and colleagues' 2021 geometric deep learning. The synthesis is organized around three themes: recursion, in which state propagation over nodes founded learning on graphs; convolution, in which spectral theory and message passing gave the graph a deep architecture; and geometry, in which attention, pooling, explainability, and symmetry made the network general. It is concluded that the GNN is deep learning's relational settlement---convolution's invariance learned from graph geometry rather than grid regularity---and that its message-passing abstraction is one of machine learning's cleanest unifications.

DOI
View source

Similar papers

#graph neural networks Oct 2026

Locality-Preserving Graph Laplacian Manifold Learning Based Model Predictive Control for Three-Phase Inverters

This article presents a model predictive control (MPC) strategy for three-phase inverters based on locality preserving projections (LPPs). Unlike conventional machine learning–based MPC approaches that rely on predefined or high-dimensional input features, the proposed LPP-MPC automatically extracts compact, informative representations by preserving the data’s intrinsic geometric structure. This dimensionality reduction enables fast linear control-law evaluation with computational complexity O(1), making the controller well-suited for real-time implementation. Experimental results demonstrate that the LPP-MPC achieves lower total harmonic distortion (THD) and reduced tracking error compared to quadratic-programming MPC under both linear and nonlinear load conditions, and the proposed controller maintains consistently lower THD throughout load transients than other methods such as two-degree-of-freedom MPC. Compared to existing model-free MPC and deep learning neural network, the LPP-MPC has the lowest THD and root mean square error with the least computational time owing to its efficient linear structure and strong generalization capability.

Jianwu Zeng, Lizheng Cheng, V. Winstead et al. · 1 citation

Related blog posts