Self-organizing maps (SOMs) are self-organized projections of high dimensional data onto a low, typically two dimensional (2D), map wherein vector similarity is implicitly translated into topological closeness in the 2D projection. They are thus used for clustering and visualization of high dimensional data. However it is often challenging to interpret the results due to drawbacks of currently used methods for identifying and visualizing cluster boundaries in the resulting feature maps. In this thesis we introduce a new phase to the SOM that we refer to as the Cluster Reinforcement (CR) phase. The CR phase amplifies within-cluster similarity with the consequence that cluster boundaries become much more evident. We also define a new Boundary (B) matrix that makes cluster boundaries easy to visualize, can be thresholded at various levels to make cluster hierarchies apparent, and can be overlain directly onto maps of component planes (something that was not possible with previous methods). The combination of the SOM, CR phase and B-matrix comprise an automated method for improved identification and informative visualization of clusters in high dimensional data. We demonstrate these methods on three data sets: the classic 13- dimensional binary-valued “animal” benchmark test, actual 60-dimensional binaryvalued phonetic word clustering problem, and 3-dimensional real-valued geographic data clustering related to fuel efficiency of vehicle choice.

Objektų klasterizavimas – tai viena iš duomenų gavybos (angl. data mining) sričių. Šių algoritmų pagrindinis privalumas – gebėjimas atpažinti grupavimo struktūrą be jokios išankstinės informacijos. Magistriniame darbe yra pristatomas vaizdų klasterizavimo algoritmas, naudojantis savaime susitvarkančius neuroninius tinklus (angl. Self-Organizing Map). Darbe analizuojami vaizdų apdorojimo, ypatingųjų taškų radimo bei palyginimo metodai. Nustatyta, kad SIFT (angl. Scale Invariant Feature Transform) ypatingųjų taškų radimas bei aprašymas veikia patikimiausiai, todėl būtent SIFT taškiniai požymiai yra naudojami klasterizavime. Darbe taip pat analizuojamas atstumo tarp paveikslėlių radimo algoritmas, tiriami skirtingi jo parametrai. Algoritmų palyginimui yra naudojamos ROC (angl. Receiver Operating Characteristic) kreivės ir EER (angl. Equal Error Rate) rodiklis. Vaizdų klasterizavimui yra naudojamas ESOM (Emergent Self-Organizing Map) neuroninis tinklas, jis vizualizuojamas U-Matrix (angl. Unified distance Matrix) pagalba ir tinklo neuronai skirstomi į klasterius vandenskyros algoritmu su skirtingu aukščio parinkimu. Magistriniame darbe demonstruojami klasterizavimo rezultatai su pavyzdinėmis paveikslėlių duomenų bazėmis bei realiais gyvenimiškais vaizdais. / Clustering algorithms – a field of data mining – aims at finding a grouping structure in the input data without any a-priori information. The master thesis is dedicated for image processing and clustering algorithms. There are point-feature detection, description and comparison methods analyzed in this paper. The SIFT (Scale Invariant Feature Transform) by D. Lowe has been shown to behave better than the other ones; hence it has been used for image to image distance calculation and undirectly in clustering phase. Finding distances between images is not a trivial task and it also has been analysed in this thesis. Several methods have been compared using ROC (Receiver Operating Curve) and EER measurements. Image clustering process is described as: (1) training of ESOM (Emergent Self-Organizing Map), (2) its visualization in U-Matrix, (3) neuron clustering using waterflood algorithm, and (4) image grouping according to their best-matching unit neurons. The paper demonstrates the image clustering algorithm on public object image databases and real life images from the Internet as well.