Trying to cluster documents using spectral clustering whether I can use SimRank measure to find the similarity between the documents?
I think you can work as usual, by clustering (K-means, e.g.) the eigenvector matrix associated with any symmetric with non negative elements matrix. Having said that, I'd recommend to work with a kernel spectral clustering approach that could take into account both SimRank results (as value at the nodes) and the original graph structure. If you want, you can find further details and bibliography of this spectral clustering adaptation in chapters 3 and 7 of my thesis.
Hope this helps.
Thesis Improving water network management by efficient division int...
Hello, I want to introduce a paper as follows.I think that it must be useful for you.
M. Yousefnezhad, D. Zhang, “Weighted Spectral Cluster Ensemble”, IEEE International Conference on Data Mining series (ICDM’15), Atlantic City, New Jersey, USA.
Conference Paper Weighted Spectral Cluster Ensemble
Yes can, but this kind of work doing the the clustering without clusters centroids
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA