Advanced Visualization Techniques for Self-organizing Maps with Graph-Based Methods

Abstract

Abstract-Self-Organizing Maps are a prominent tool for exploratory data analysis. In this paper, we propose a method of visualizing the cluster structure of the SOM based on the similarity of areas on the map, computed by aggregation of the distances of the underlying component planes of the codebook. The result can then be plotted on top of the map lattice with arrows that point to the closest cluster center, which is analogous to flow and vector field visualizations. A parameter is provided that allows fine-tuning of the granularity of the clustering, which can be adjusted according to whether a global or local view on the map is desired. We provide experimental results with a real-world data set where we discuss the effects of parametrization and the general applicability of our method, along with comparison to related techniques.

Data and dimension reduction for visual financial performance analysis

“…73 Examples of aids for assessing distance structures are Sammon's mapping, the Unified distance matrix (U-matrix), 74 and cluster connections. 75 Moreover, some methods attempt to account for both structures and densities, such as the U*-matrix, 76 the sky metaphor visualization, 77 the neighborhood graph, 78 and smoothed data histograms. 79 The second group consists of visualizations that enhance the representation of multidimensional information.…”

Exploring and Understanding the High Dimensional and Sparse Image Face Space: a Self-Organized Manifold Mapping

“…A simple way to explain this algorithm is to understand the output neurons, represented by the weights ij w computed in the SOM algorithm, as a set of nodes of a fully connected graph (Cormen et al 2001;Pölzlbauer, Rauber, Dittenbach, 2005;Mayer, Rauber, 2010 ) in the parameter space. Each edge in this graph has a cost given by the Euclidean distance between its ends.…”

A SOM-Based Technique for a User-Centric Content Extraction and Classification of Web 2.0 with a Special Consideration of Security Aspects

“…Vector Activity Histogram, Class Visualization [15], Component Planes [27], Vector Fields [20], Hit Histogram, Metro Map [17], Minimum Spanning Tree, Neighborhood Graph [21], Smoothed Data Histograms [19], Sky Metaphor Visualization [14], U-Matrix [26], D-Matrix, P-Matrix [24] and U * -Matrix [25]. Furthermore as an additional task we have to take the extra information of Web 2.0 entries into consideration throughout the document processing phase.…”