CMStatistics 2018: Start Registration
View Submission - CMStatistics
Title: Computational efficiency for fuzzy clustering Authors:  Takatsugu Yoshioka - Okayama University of Science (Japan)
Masahiro Kuroda - Okayama University of Science (Japan)
Yuichi Mori - Okayama University of Science (Japan) [presenting]
Abstract: One of the most widely used soft clustering algorithms is the Fuzzy c-means clustering (FCM), which searches a reasonable form of clustering in which each data point belongs to multiple clusters. FCM therefore requires high computational cost due to the iterative computation to estimate two parameters (membership and cluster centroid matrices) alternately. Because this computational algorithm is a kind of alternating least squares (ALS) algorithm, so a general procedure to accelerate ALS type of iteration using the vector epsilon algorithm can be applied to FCM computation to obtain the computational results faster than the original computation. The performance/efficiency of the vector epsilon accelerated FCM algorithm is evaluated in simulations under several conditions e.g., data size, the number of original clusters, the number of estimated clusters, accelerated parameter (membership or centroid) and so on. These numerical experiments demonstrate that the vector epsilon accelerated FCM accelerates the computation twice or more as fast as the original one.