Keywords: Predictive models, non-linear, neural network, least square, joint inversion.
This study focuses on the effectiveness of combining data sets from different predictive models that share some common parameter sets. Additional sets of data from different predictive models can be considered as a priori information contaminated with arbitrary errors. Predictive models may employ a simple linear equation, a non-linear neural network or a combination of models of different complexity. The effectiveness of combining data types is understood by analyzing operators from a non-linear least square joint inversion system. We will show results from idealized non-linear models.