Abstract | ||
---|---|---|
This paper provides the theory and the machinery for the generalization of the celebrated mean-shift algorithm to exponential families. We show that the baseline version of the algorithm is a special case of the proposed one, the one formed by the multivariate normal exponential family with known covariance matrix. With the proposed generalization, we will be capable of clustering entities that lie on other probabilistic manifolds, and hence to increasing its applicability significantly. An example is given for the problem of speaker clustering. |
Year | DOI | Venue |
---|---|---|
2012 | 10.1109/ISSPA.2012.6310605 | 2012 11th International Conference on Information Science, Signal Processing and their Applications (ISSPA) |
Keywords | Field | DocType |
mean shift algorithm,exponential family manifolds,multivariate normal exponential family,covariance matrix,clustering entities,probabilistic manifolds,speaker clustering | Kernel (linear algebra),Pattern recognition,Exponential family,Multivariate normal distribution,Artificial intelligence,Probabilistic logic,Covariance matrix,Cluster analysis,Manifold,Mathematics,Special case | Conference |
ISBN | Citations | PageRank |
978-1-4673-0381-1 | 0 | 0.34 |
References | Authors | |
11 | 4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Themos Stafylakis | 1 | 431 | 30.12 |
Vassilios Katsouros | 2 | 73 | 10.63 |
Patrick Kenny | 3 | 2700 | 214.80 |
Pierre Dumouchel | 4 | 1759 | 129.78 |