A Probabilistic Framework for Graph Clustering

Abstract

The paper describes a probabilistic framework for graph clustering. We commence from a set of pairwise distances between graph structures. From this set of distances, we use a mixture model to characterize the pairwise affinity of the different graphs. We present an EM-like algorithm for clustering the graphs by iteratively updating the elements of the affinity matrix. In the M-step we apply eigendcomposition to the affinity matrix to locate the principal clusters. In the M-step we update the affinity probabilities. We apply the resulting unsupervised clustering algorithm to two practical problems. The first of these involves locating shape-categories using shock trees extracted from 2D silhouettes. The second problem involves finding the view structure of a polyhedral object using the Delaunay triangulation of corner features.

Cite

Text

Luo et al. "A Probabilistic Framework for Graph Clustering." IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2001. doi:10.1109/CVPR.2001.990621

Markdown

[Luo et al. "A Probabilistic Framework for Graph Clustering." IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2001.](https://mlanthology.org/cvpr/2001/luo2001cvpr-probabilistic/) doi:10.1109/CVPR.2001.990621

BibTeX

@inproceedings{luo2001cvpr-probabilistic,
  title     = {{A Probabilistic Framework for Graph Clustering}},
  author    = {Luo, Bin and Robles-Kelly, Antonio and Torsello, Andrea and Wilson, Richard C. and Hancock, Edwin R.},
  booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  year      = {2001},
  pages     = {I:912-919},
  doi       = {10.1109/CVPR.2001.990621},
  url       = {https://mlanthology.org/cvpr/2001/luo2001cvpr-probabilistic/}
}