Transactions on Machine Learning and Data Mining (ISSN: 1865-6781)
Volume 2 - Number 2 - October 2009 - Page 80-99
Clustering and Separating of a Set of Members in Terms of Mutual Distances and Similarities
S. D. Dvoenko
Tula State University, 300600, Tula, Russia
In a case of set members are presented via mutual distances or similarities well-known algorithms for clustering ( K-means), grouping (Modulus), and learning (Kozinets's) are under investigation. Relationship between K-means and Modulus algorithms is shown based on idea of unbiased partitioning. The problem of learning to recognize set members (objects or features) is under investigation too. Experimental results are shown for feature recognition (Holzinger's psychological tests) and for object recognition (small classes of amino-acid sequences) problems.
Download Paper (308 KB)