This chapter illustrates the application of Kohonen self‐organizing maps (Kohonen SOMs) to the visualization and classification of multivariate data using a set of molecules assigned to five types of biological activity. Typical applications of Kohonen SOMs in chemoinformatics have been the classification of molecules according to biological activity, selection of data sets encompassing large diversity, identification of redundant features (with transposed features' matrices), and generation of molecular descriptors by mapping of molecular components (e.g., bonds and points of the molecular surface). The SOM algorithm (explained below) is typically based on Euclidean distances between neurons and objects represented by descriptors. Therefore, the numerical range of a descriptor can influence its impact on the mapping. Normalization of the data is thus required previous to training. Each neuron of a Kohonen SOM contains as many elements (weights) as the number of input features for the objects to be mapped.