ИСТИНА |
Войти в систему Регистрация |
|
ИСТИНА ИНХС РАН |
||
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.