Kohonen feature maps

Self-organizing maps (also called SOMs, Kohonen feature maps, or kmaps) are special kinds of artificial neural networks (ANNs) that are able to represent sets of descriptors in a low-dimensional map [114—116], and are increasingly applied for mapping of various molecular data in the fields of analytical chemistry and drug design [89, 117, 118]. 

T. Kohonen, Self Organization and Associated Memory. Springer-Verlag, Heidelberg, 1989. W.J. Meissen, J.R.M. Smits, L.M.C. Buydens and G. Kateman, Using artificial neural networks for solving chemical problems. II. Kohonen self-organizing feature maps and Hopfield networks. Chemom. Intell. Lab. Syst., 23 (1994) 267-291. 

In contrast to common ANNs, Kohonen networks produce self-organized topological feature maps (Kohonen [1982, 1984]). The basic idea of Kohonen mapping is that information in data usually contains not only an algebraic but also a topological aspect. These double aspect is shown schematically in Fig. 8.25 where the data and the structure of them are composed. 

Kohonen T (1982) Self-organized formation of topologically correct feature maps. Biol Cybern 43 59... 

Also known as a Self-Organizing Feature Map or SOFM, or a Kohonen map after its inventor. 

The utility of ANNs as a pattern recognition technique in the field of microbeam analysis was demonstrated by Ro and Linton [99]. Back-propagation neural networks were applied to laser microprobe mass spectra (LAMMS) to determine interparticle variations in molecular components. Selforganizing feature maps (Kohonen neural networks) were employed to extract information on molecular distributions within environmental microparticles imaged in cross-section using SIMS. 

A specialized method for similarity-based visualization of high-dimensional data is formed by self-organizing feature maps (SOM). The data items are arranged on a two-dimensional plane with the aid of neural networks, especially Kohonen nets. Similarity between data items is represented by spacial closeness, while large distances indicate major dissimilarities [968]. At the authors department, a system called MIDAS had already been developed which combines strategies for the creation of feature maps with the supervised generation of fuzzy-terms from the maps [967]. 

Several QSAR approaches are based on Kohonen maps, such as topological feature maps. Comparative Molecular Surface Analysis, and MOLMAP descriptors. Counter-propagation neural network is a development of Kohonen maps for classification purposes [Zupan, Novic et al., 1995], which considers a set of output layers, called Grosberg... 

Kohonen networks are also termed self-organizing nets or self-organizing feature maps. 

Kohonen, T. A Simple Paradigm For The Self-Organized Formatiom Of Structured Feature Maps In Competition And Cooperation In Neural Nets (Lecture notes in biomathematics vol 45, Amari, S. Arbib, M. A. Eds.) ISBN 0387115749 Springer-Verlag Berlin, 1982. 

T. Kohonen, K. Masisara, and T. Saramaki, Phonotopic Maps—Insightful Representation of Phonological Features for Speech Representation, Proceedings IEEE 7th International Conference, Montreal, Canada (1984). 

Input/Output sequence encoding methods (see Table 9.1) PRF direct encoding of residue profile a = a-helix fi = P-sheet C = random coil t = (5-reverse turn L = loop %AA = amino acid composition FEAT = indirect encoding of sequence features MF = memory factor dist = distance Map = Kohonen map (with dimensions). 

MOLMAP (Molecular Map of Atom-level Properties) descriptors are uniform-length vectorial descriptors derived by mapping physico-chemical properties of all the bonds in a molecule into a 2D Kohonen —> self organizing map (SOM) [Zhang and Aires-de-Sousa, 2005 Gupta, Metthew ef al., 2006]. These descriptors encode local features of a chemical structure, being calculated on the basis of properties of single elements in a molecule, such as bonds. 

---