Applying artificial neural networks to solve the inverse problem of evaluating concentrations in multianalyte mixtures from biosensor signals