Automated line search in molecular spectra based on nonparametric statistical methods: Regularization in estimating parameters of spectral lines

The application of pattern-recognition methods to automated processing of spectroscopic information is described. Model shapes of spectral lines are fitted by the least squares method with a regularizing addition. Some additional ways of increasing the stability of the fitting process are described. The results can be interesting for specialists who solve problems of spectroscopy and pattern recognition and construct expert and developmental support information systems.