Things you haven't heard about the self-organizing map In Neural Networks, 1993., IEEE International Conference on (1993), pp. 1147-1156 vol.3, doi:10.1109/icnn.1993.298719 by T. Kohonen

@inproceedings{kohonen-1993, abstract = {The self-organizing map ({SOM}) algorithm can be related to a biological neural network in many essential known details; even cyclic behavior automatically ensues from a simple nonlinear neural model, whereby these cycles correspond to the steps of the discrete-time {SOM} algorithm. Compared with the other traditional neural-network algorithms, the {SOM} alone has the advantage of tolerating very low accuracy in the representation of its signals and synaptic weights. This is proven by simulations. Such a property ought to be shared by any realistic neural-network model. While the {SOM} can thus be advanced as a genuine neural-network paradigm, it is shown how the basic algorithm can be generalized and made more computationally efficient in several ways}, author = {Kohonen, T.}, booktitle = {Neural Networks, 1993., IEEE International Conference on}, doi = {10.1109/icnn.1993.298719}, institution = {Lab. of Comput. \& Inf. Sci., Helsinki Univ. of Technol., Espoo}, isbn = {0-7803-0999-5}, keywords = {ann, biology, learning, math, som, unsupervised-learning}, pages = {1147--1156 vol.3}, posted-at = {2011-12-12 10:57:49}, priority = {2}, publisher = {IEEE}, title = {Things you haven't heard about the self-organizing map}, url = {http://dx.doi.org/10.1109/icnn.1993.298719}, year = {1993} }

See the CiteULike entry for more info, PDF links, BibTex etc.