Test series selection from nonlinear neural mapping

A new nonlinear neural mapping (N2M) technique based on the combined use of Kohonen self-organizing map (KSOM), minimum spanning tree (MST), and nonlinear mapping (NLM) is introduced for optimal test series selection. With the N2M method, KSOM results are enhanced by the visualization of the actual distances between the loaded neurons from MST and NLM. N2M provides an easily interpretable and comprehensible graphical display which guides the selection of representative test series especially when the number of individuals is high. In addition, structure-activity relationships can be derived. The approach is open since any information useful for data interpretation can be plotted by means of graphical tools.