ИСТИНА

When a multi-parameter inverse problem is solved with artificial neural networks, it is usually solved separately for each determined parameter (autonomous determination). In their preceding studies, the authors have demonstrated that joining parameters into groups with simultaneous determination of the values of all parameters within each group may in some cases improve the precision of solution of inverse problems. In this study, the observed effect has been investigated in respect to its resistance to noise in data. The study has been performed at the example of the inverse problem of magnetotellurics, which has a high dimensionality.