Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Nonlinear Higher-Order Label Spreading

About

Label spreading is a general technique for semi-supervised learning with point cloud or network data, which can be interpreted as a diffusion of labels on a graph. While there are many variants of label spreading, nearly all of them are linear models, where the incoming information to a node is a weighted sum of information from neighboring nodes. Here, we add nonlinearity to label spreading through nonlinear functions of higher-order structure in the graph, namely triangles in the graph. For a broad class of nonlinear functions, we prove convergence of our nonlinear higher-order label spreading algorithm to the global solution of a constrained semi-supervised loss function. We demonstrate the efficiency and efficacy of our approach on a variety of point cloud and network datasets, where the nonlinear higher-order model compares favorably to classical label spreading, as well as hypergraph models and graph neural networks.

Francesco Tudisco, Austin R. Benson, Konstantin Prokopchik• 2020

Related benchmarks

TaskDatasetResultRank
Node ClassificationCora
Accuracy78.48
1215
Node ClassificationCora (test)
Mean Accuracy78.48
861
Node ClassificationCiteseer (test)
Accuracy0.7521
824
Node ClassificationChameleon
Accuracy44.95
640
Node ClassificationSquirrel
Accuracy40.13
591
Node ClassificationChameleon (test)
Mean Accuracy44.95
297
Node ClassificationCornell (test)
Mean Accuracy75.14
274
Node ClassificationTexas (test)
Mean Accuracy83.51
269
Node ClassificationSquirrel (test)
Mean Accuracy40.13
267
Node ClassificationWisconsin (test)
Mean Accuracy86.67
239
Showing 10 of 31 rows

Other info

Follow for update