Neural Network Solution of an Inverse Problem in Raman Spectroscopy of Multi-Component Solutions of Inorganic Salts: Group Determination as a Method to Increase Noise Resilience of the Solution⁎

Many inverse problems are incorrect or ill-conditioned. This makes them very sensitive to noise in the input data. In the previous studies of the authors, it was shown that under certain conditions simultaneous determination of several parameters allows increasing the resilience of the neural network solution to noise in the input data. In the present study, applicability of this approach to the inverse problem of spectroscopy, which consists of determining the concentrations of 10 ions in a multi-component solution of inorganic salts by Raman spectra, is demonstrated.