Utilizing the Topology Preserving Property of Self-organizing Maps for Classiication
نویسندگان
چکیده
The Kohonen Self-Organizing Map (SOM) is a popular algorithm for constructing a nearest neighbor codebook in pattern space. The algorithm utilizes a prede ned ordering on the codebook to distribute the codes proportionally on the input manifold. In the end this ordering should re ect the structure of the input. Prototypical application of the SOM uses the codebook but neglects the ordering. We explore the practical possibilities for taking advantage of the ordering, concentrating mainly on classi cation tasks. We present three approaches: coding class boundaries with a duo of SOMs, construction of radial basis function networks with ordering information and using a SOM as a preprocessor for backpropagation networks. We obtain positive results on a number of real world data sets from the eld of medical diagnosis, speech{ and image processing. From this we conclude the ordering property of SOMs contains useful information. However, it is still unclear how to pro t from it in the best possible way.
منابع مشابه
Mnemonic SOMs: Recognizable Shapes for Self-Organizing Maps
The Self-Organizing Map (SOM) enjoys significant popularity in the field of data mining and visualization. While its topology-preserving mapping allows easier interpretation of complex data, communicating the location of clusters and individual data items as well as memorizing locations are not solved satisfactorily in conventional rectangular maps. In this paper, a variant of self-organizing m...
متن کاملParallel Manner and Twist Measurement for Self-Organizing Maps
This paper investigates a self-organizing map algorithm which is simultaneously given multiple inputs, and introduces a criterion of ordering process as the twist measurement for reference vectors in multidimensional array. The self-organizing algorithm and the criterion are termed the parallel self-organizing algorithm and the index of twist, respectively. The algorithm updates reference vecto...
متن کاملCodebook Clustering by Self-organizing Maps for Fractal Image Compression
A fast encoding scheme for fractal image compression is presented. The method uses a clustering algorithm based on Kohonen's self-organizing maps. Domain blocks are clustered, yielding a classiication with a notion of distance which is not given in traditional classiication schemes.
متن کاملOral Presentations Self organizing map and counterpropagation neural networks in structure-property modelling: Examples from environmental science and drug design
Self organizing map technique is often used to analyse the data in multi-dimensional space. Basis of this technique is a non-linear projection from multi-dimensional space onto two-dimensional grid of neurons (map). The projection, which is topology preserving but not metric preserving, is achieved via non-linear algorithm known as training. The fundamental property of the trained network is th...
متن کاملA combined measure for quantifying and qualifying the topology preservation of growing self-organizing maps
Keywords: Topology preserving Self-organizing map Growing cell structures Visualization methods Delaunay triangulation The Self-Organizing Map (SOM) is a neural network model that performs an ordered projection of a high dimensional input space in a low-dimensional topological structure. The process in which such mapping is formed is defined by the SOM algorithm, which is a competitive, unsuper...
متن کامل