SPECSOM

For a given chemical system the shape of the measured spectra depends on physico-chemical parameters, i.e. temperature, fractions of compounds, pH values, etc. This complicated high dimensional data space, which consists of spectra and their corresponding physico-chemical parameters, can be unscrambled by the application of a self-organizing feature map (SOM) as a type of an artificial neural network. Our SPECSOM yields a discretized two-dimensional view of high dimensional data space after its training phase by using the spectral data. Thus, SPECSOM enables an eased study of interrelationships between the spectra and the physico-chemical parameters. Moreover, SPECSOM can be used for the prediction of spectra and/or physico-chemical parameters in the case of sparse data, i.e. for a given spectrum the corresponding unknown set of physico-chemical parameters can be predicted and vice versa.