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

Beyond Spectral Clustering: Probabilistic Cuts for Differentiable Graph Partitioning

About

Probabilistic relaxations of graph cuts offer a differentiable alternative to spectral clustering, enabling end-to-end and online learning without eigendecompositions, yet prior work centered on RatioCut and lacked general guarantees and principled gradients. We present a unified probabilistic framework that covers a wide class of cuts, including Normalized Cut. Our framework provides tight analytic upper bounds on expected discrete cuts via integral representations and Gauss hypergeometric functions with closed-form forward and backward. Together, these results deliver a rigorous, numerically stable foundation for scalable, differentiable graph partitioning covering a wide range of clustering and contrastive learning objectives.

Ayoub Ghriss• 2025

Related benchmarks

TaskDatasetResultRank
Image ClusteringCIFAR-10
NMI0.878
318
ClusteringFashion MNIST
NMI76
107
Image ClusteringDTD
NMI69.7
49
ClusteringSTL-10
ACC99.8
28
ClusteringCIFAR-100
NMI87.3
25
ClusteringPets
NMI93.7
21
ClusteringCIFAR-10
ACC98.6
15
ClusteringImagenette
Accuracy (%)99.8
12
ClusteringEuroSAT
Accuracy93.2
12
ClusteringFlowers-102
Accuracy99.7
12
Showing 10 of 17 rows

Other info

Follow for update