The nature of statistical learning theory
The nature of statistical learning theory
Comparison of the performance of center-based clustering algorithms
PAKDD'03 Proceedings of the 7th Pacific-Asia conference on Advances in knowledge discovery and data mining
The mahalanobis distance based rival penalized competitive learning algorithm
ISNN'06 Proceedings of the Third international conference on Advances in Neural Networks - Volume Part I
| Hi-index | 0.00 |
This paper presents an alternative for center-based clustering algorithms, in particular the k-means algorithm, via statistical learning analysis. The essence of statistical learning principle, i.e., both the empirical risk and structural assessment, is taken into particular consideration for the clustering algorithm so as to derive and develop the relevant minimization mathematical criterion with automatic parameter learning and model selection in parallel. The proposed algorithm roughly decides on the number of clusters, by earning activation for the winners and assigning penalty for the rivals, so that the most competitive center wins for possible prediction and the extra ones are driven far away when starting the algorithm from a too large number of clusters without any prior knowledge. Simulation experiments prove the feasibility of the algorithm and show good performances of the double learning tasks during clustering.