References: Label Propagation and Semi-Supervised Learning¶
-
Label Propagation Algorithm - Wikipedia - Covers the family of iterative label propagation methods on graphs, including the Zhu & Ghahramani harmonic function formulation and the Zhou et al. label spreading variant, with discussion of convergence conditions.
-
Belief Propagation - Wikipedia - Explains the sum-product and max-product message-passing algorithms on factor graphs, including exact inference on trees and the loopy BP approximation used when graphs contain cycles.
-
SIR Model - Wikipedia - Describes the Susceptible-Infected-Recovered compartmental epidemic model, its differential equation formulation, the basic reproduction number R₀, and extensions to network-structured populations.
-
Semi-Supervised Learning - Chapelle, O., Schölkopf, B., & Zien, A. (Eds.) - MIT Press - Comprehensive edited volume covering graph-based methods, manifold regularization, transductive SVMs, and generative approaches; Chapter 11 gives the definitive treatment of Gaussian fields and label propagation.
-
Networks, Crowds, and Markets: Reasoning About a Highly Connected World - Easley, D. & Kleinberg, J. - Cambridge University Press - Rigorous undergraduate-accessible text unifying game theory, economics, and network science; Chapters 19–21 cover diffusion, cascades, and epidemic models with the same formalism used in this chapter.
-
Semi-supervised Classification with Graph Convolutional Networks (Kipf & Welling, 2017) - arXiv - Introduces the GCN spectral convolution and frames node classification as a semi-supervised problem; the label propagation interpretation of GCN motivates the direct connection between this chapter and GNN foundations.
-
Predict then Propagate: Graph Neural Networks meet Personalized PageRank (Gasteiger et al., 2019) - arXiv - Derives APPNP by replacing GCN aggregation with personalized PageRank diffusion, showing that separating feature transformation from label propagation improves both depth and accuracy.
-
Combining Label Propagation and Simple Models Out-performs Graph Neural Networks (Huang et al., 2021) - arXiv - Introduces the Correct & Smooth (C&S) post-processing framework that applies two rounds of label diffusion on top of a plain MLP, achieving state-of-the-art results on ogbn-arxiv with minimal compute.
-
Node Classification — PyTorch Geometric Documentation - PyTorch Geometric Docs - Step-by-step tutorial for semi-supervised node classification using GCN and related models on the Cora dataset, with runnable code that bridges the label propagation ideas in this chapter to practical GNN implementations.
-
Papers With Code — Semi-Supervised Node Classification - Papers With Code - Tracks state-of-the-art results and links to reproducible implementations for semi-supervised node classification benchmarks including Cora, CiteSeer, and ogbn-arxiv, making it easy to compare label propagation baselines against current GNN methods.