June 13, 2018, dr inż. Piotr Lasek, University of Rzeszów, “Granular computing in clustering”
Abstract: The key-element of each clustering algorithm is determining distances between clustered objects. In order to make this process efficient various indexes are used. In our presentation we will focus on a modification of the k-means algorithm employing so-called pi-cube – an inductive hierarchical structure representing data on different levels of granularity. Under certain conditions, the modification of the algorithm makes determining clusters more efficient due to the fact that the whole data set does not have to be scanned when assigning points to clusters.