Arthur Flexer. On the use of self-organizing maps for clustering and visualization. In PKDD'99, pages 80--88, 1999.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....space while preserving as much as possible the structure of the data in the high dimensional data space. This is achieved by mapping points in one space to points in another space such that nearby points map to nearby points (and sometimes in addition far away points map to far away points) [11]. Galaxies [32, 31] visualization displays clusters and document interrelatedness by reducing a high dimensional representation of documents to a two dimensional scatterplot. The documents are clustered in the high dimensional space through a metric of similarity such as Euclidean distance or ....

....category map. Usually interrelated words that have similar context appear close to each other on the map. Then the documents are encoded by mapping their text onto the word category map. The document map is then formed with a SOM algorithm using the document vectors in word category map space. In [11], the use of self organizing maps for clustering and visualization was discussed in depth. A comparative study on the quality and effectiveness of SOMs and Sammon s mapping when applying to classification and visualization was reported. A number of other dimension reduction algorithms have been ....

A. Flexer. On the use of self-organizing maps for clustering and visualization. Intelligent-Data-Analysis, 5:373--84, 2001.

....none of the existing multidimensional visualization techniques can map all the dimensions at the same time without cluttering the display. Popular dimension reduction approaches, such as Principal Component Analysis [14] Multidimensional Scaling [18] and Kohonen s Self Organizing Maps [16, 10], condense the hundreds or thousands of dimensions into a few dimensions. However, those generated dimensions have little intuitive meaning to users and allow little user interaction. Dimension filtering is more intuitive to users in that the remaining dimensions are all original dimensions in the ....

A. Flexer. On the use of self-organizing maps for clustering and visualization. PKDD'99, p. 80-88, 1999.

....of the input distribution and satisfaction of user s constraints on the category representation, without losing the clustering quality of ART. B. Reuters 21578 Corpus Our experiments on the Reuters 21578 corpus compared the performance of ART C with those of Self Organizing Map (SOM) 8][9] and k means [10] both have been extensively studied in the literature. The major di#erences among the trio can be summarized as below: SOM performs incremental and soft learning with relatively low learning rate that requires a large number of learning iterations; k means is an ....

A. Flexer, "On the use of self-organizing maps for clustering and visualization," in Principles of Data Mining and Knowledge Discovery, 1999, pp. 80--88.

....it carries, and then visualize the data set in the reduced dimensional space. There are several popular dimensionality reduction techniques used in data visualization, including Principal Component Analysis (PCA) 13] Multidimensional Scaling (MDS) 17] and Kohonen s Self Organizing Maps (SOM) [16, 6]. These approaches have a major drawback in that the generated low dimensional subspace has no intuitive meaning to users. In addition, little user interaction is generally allowed in those highly automatic processes, thus users have difficulty applying their domain acknowledge to improve the ....

....to project data down to a few dimensions that account for most of the variance within the data. Multidimensional Scaling (MDS) 17] is an iterative non linear optimization algorithm for projecting multidimensional data down to a reduced number of dimensions. Kohonen s Self Organizing Maps (SOM) [16, 6] is an unsupervised learning method for reducing multidimensional data to 2D feature maps [3] There are many visualization systems that make use of existing dimensionality reduction techniques [23, 3, 11] Galaxies and ThemeScape [23] project high dimensional document vectors and their cluster ....

A. Flexer. On the use of self-organizing maps for clustering and visualization. PKDD'99, p. 80-88, 1999.

....correlation between the distances of points in the input space and the distances of their projections in the two dimensional space. BN95] found that the traditional statistical methods preserve the distances much more effectively than the SOM algorithm. This result was also confirmed by Flexer [Fle99]. In an empirical study on SOM s ability to do both clustering and topology preservation in the same time, Flexer [Fle97, Fle99] compared the SOM algorithm to a combined technique of k means clustering plus Sammon mapping [Sam69] a traditional statistical method used for multidimensional ....

....space. BN95] found that the traditional statistical methods preserve the distances much more effectively than the SOM algorithm. This result was also confirmed by Flexer [Fle99] In an empirical study on SOM s ability to do both clustering and topology preservation in the same time, Flexer [Fle97, Fle99] compared the SOM algorithm to a combined technique of k means clustering plus Sammon mapping [Sam69] a traditional statistical method used for multidimensional scaling) on the cluster centers. If zero neighborhood width was used in the final iterations of the SOM algorithm, the SOM algorithm ....

A Flexer. On the use of self-organizing maps for clustering and visualization. In J.M. Zytkow and J. Rauch, editors, Principles of Data Mining and Knowledge Discovery, Third European Conference, PKDD'99, Lecture Notes in Artificial Intelligence 1704, pages 80--88, Prague, Czech Republic, 1999. Springer. 114

....by an ordering of the cluster centroids in a one or two dimensional output space. The latter property is closely related to multidimensional scaling (MDS) in statistics. More on SOMs relation to clustering and MDS can be found in [Balakrishnan et al. 1994] Flexer 1997] Bishop et al. 1998] and [Flexer 1999]. Using the Kalman coefficients for every one second segment as input vectors for the SOM neglecting their ordering in time, the authors observed eight clearly visible clusters in the output map. Even more important they were able to distinguish three typical trajectories between these clusters ....

Flexer A.: On the use of self-organizing maps for clustering and visualization, in Zytkow J.M. & Rauch J.(eds.), Principles of Data Mining and Knowledge Discovery, Third European Conference, PKDD'99, Prague, Czech Republic, Proceedings, Lecture Notes in Artificial Intelligence 1704, Springer, p.80-88, 1999.

No context found.

Arthur Flexer. On the use of self-organizing maps for clustering and visualization. In PKDD'99, pages 80--88, 1999.

No context found.

FLEXER, A. 1999. On the use of self-organizing maps for clustering and visualization. In Principles of Data Mining and Knowledge Discovery, 80--88.

No context found.

A. Flexer, On the use of self-organizing maps for clustering and visualization, Intelligent Data Analysis 5 (2001), no. 5, 373--384.

No context found.

Flexer, A. (1999). On the use of self-organizing maps for clustering and visualization. Principles of Data Mining and Knowledge Discovery. Z. J.M. and R. J., Springer. 1704: 80-88.

No context found.

Flexer, A. (2001). On the use of self-organizing maps for clustering and visualization. Intelligent Data Analysis, 5:373--84.

No context found.

Flexer, A. (1999). On the use of self-organizing maps for clustering and visualization. In Principles of Data Mining and Knowledge Discovery, pages 80--88.

No context found.

Flexer, A. (1999). On the use of self-organizing maps for clustering and visualization. In Proc. of the 3rd Europ. Conf. on Principles of Data Mining and Knowl. Discovery, volume 1704 of LNAI, pages 80--88, Prague, Czech Republic. Springer.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC