Name
Abstract | ||
|---|---|---|
Models for large, sparse graphs are found in many applications and are an active topic in machine learning research. We develop a new generative model that combines rich block structure and simple, efficient estimation by collapsed Gibbs sampling. Novel in our method is that we may learn the strength of assortative and disassortative mixing schemes of communities. Most earlier approaches, both based on low-dimensional projections and Latent Dirichlet Allocation implicitely rely on one of the two assumptions: some algorithms define similarity based solely on connectedness while others solely on the similarity of the neighborhood, leading to undesired results for example in near-bipartite subgraphs. In our experiments we cluster both small and large graphs, involving real and generated graphs that are known to be hard to partition. Our method outperforms earlier Latent Dirichlet Allocation based models as well as spectral heuristics. |
Year | DOI | Venue |
|---|---|---|
2010 | 10.1145/1830252.1830261 | MLG@KDD |
Keywords | Field | DocType |
efficient estimation,near-bipartite subgraphs,efficient block model,new generative model,collapsed gibbs,latent variables,statistical network analysis,block model,large graph,low-dimensional projection,sparse graph,rich block structure,latent dirichlet allocation,earlier approach,active topic,hierarchical bayes,machine learning,gibbs sampling,network analysis,latent variable | Dynamic topic model,Data mining,Latent Dirichlet allocation,Computer science,Latent variable,Heuristics,Probabilistic latent semantic analysis,Artificial intelligence,Cluster analysis,Gibbs sampling,Machine learning,Generative model | Conference |
Citations | PageRank | References |
2 | 0.37 | 12 |
Authors | ||
3 | ||
Name | Order | Citations | PageRank |
|---|---|---|---|
| Ádám Gyenge | 1 | 2 | 0.71 |
| Janne Sinkkonen | 2 | 231 | 21.36 |
| András A. Benczúr | 3 | 574 | 60.14 |