A survey of some classic self-organizing maps with incremental learning
Xinji Qiang, Guojian Cheng, Zhen Li · 2010
Kohonen's Self-Organizing Maps (SOM) is a class of typical artificial neural networks (ANN) with unsupervised learning which has been widely used in clustering tasks, dimensionality reduction, data mining, information extraction, density approximation, data compression, etc. A basic principle of unsupervised learning is the competition mechanism, in which the output neurons compete for activation. In most competitive learning algorithms only one output neuron is activated at any given time. This is realized by means of the so-called winner- takes-all mode. Another mode is winner-takes-more. In this paper, the competitive learning is firstly introduced, the SOM topology and leaning mechanism are then illustrated. Thirdly, some self-organizing maps with incremental learning (SOMIL), such as self-organizing surfaces, evolve self-organizing maps, incremental grid growing and growing hierarchical self-organizing map, are outlined. Finally, the new development of SOMIL is reviewed. Some conclusions are given at the end of the paper.