kMeans is the traditional algorithm that is used in Spectral clustering? Can we apply kMeans ++, other algorithm in Spectracl clustering?
You may go with Fuzzy C-Mean Clustering. It proved very efficient than KMeans Clustering
Hi Kavitha,
of course you can plug the clustering algorithm that you prefer in this phase of spectral clustering. The wide use of k-means embedded in specc is because its simplicity and usually good results.
Hi Nagaja ka,
c-means algorithm is similar to k-means. The difference is that each element has a degree of belonging to clusters (fuzzy theory approach) instead of the deterministic point of view in the membership, when approaching the groups by k-means.
Kavitha, you can use connected component algorithm, Meila and Shi algorithm, Kannan Spectral algorithm instead of K-mean algorithm. You can refer the attached paper for the comparison among the algorithms.
Sir,
What about Kernel K Means.. In most of the papers i have found this algorithm... Can any one explain it..
Dear Kavitha,
if this question might be still interesting to you, we found that the application of k-means is actually motivated by the minimum cut objective. So in principle, you can use any clustering algorithm but if you want to minimize the ratio/normalized cut then it makes sense to use k-means. We explored this relationship and used our insights to define a more robust version of spectral clustering here:
Conference Paper The SpectACl of Nonconvex Clustering: a Spectral Approach to...
- Badges
- Science topic
- Similar topics
- Books