Variants of self-organizing maps

@article{citeulike:2588167, abstract = {Self-organizing maps have a bearing on traditional vector quantization. A characteristic that makes them more closely resemble certain biological brain maps, however, is the spatial order of their responses, which is formed in the learning process. A discussion is presented of the basic algorithms and two innovations: dynamic weighting of the input signals at each input of each cell, which improves the ordering when very different input signals are used, and definition of neighborhoods in the learning algorithm by the minimal spanning tree, which provides a far better and faster approximation of prominently structured density functions. It is cautioned that if the maps are used for pattern recognition and decision process, it is necessary to fine tune the reference vectors so that they directly define the decision borders}, author = {Kangas, J. A. and Kohonen, T. K. and Laaksonen, J. T. }, booktitle = {Neural Networks, IEEE Transactions on}, citeulike-article-id = {2588167}, doi = {10.1109/72.80208}, journal = {Neural Networks, IEEE Transactions on}, keywords = {clustering, phd}, number = {1}, pages = {93--99}, posted-at = {2008-03-26 02:45:55}, priority = {0}, title = {Variants of self-organizing maps}, url = {http://dx.doi.org/10.1109/72.80208}, volume = {1}, year = {1990} }

TY - JOUR ID - citeulike:2588167 L3 - citeulike-article-id:2588167 TI - Variants of self-organizing maps T2 - Neural Networks, IEEE Transactions on JF - Neural Networks, IEEE Transactions on VL - 1 IS - 1 SP - 93 EP - 99 N2 - Self-organizing maps have a bearing on traditional vector quantization. A characteristic that makes them more closely resemble certain biological brain maps, however, is the spatial order of their responses, which is formed in the learning process. A discussion is presented of the basic algorithms and two innovations: dynamic weighting of the input signals at each input of each cell, which improves the ordering when very different input signals are used, and definition of neighborhoods in the learning algorithm by the minimal spanning tree, which provides a far better and faster approximation of prominently structured density functions. It is cautioned that if the maps are used for pattern recognition and decision process, it is necessary to fine tune the reference vectors so that they directly define the decision borders KW - clustering KW - phd AU - Kangas, JA AU - Kohonen, TK AU - Laaksonen, JT PY - 1990/// UR - http://dx.doi.org/10.1109/72.80208 ER -