Figure 10
From: Towards an efficient validation of dynamical whole-brain models

Distributions of the optimal model parameters detected by the considered techniques across all subjects. For each approach and subject, the five best solutions featuring the highest goodness-of-fit values are considered to estimate the distribution. They were selected from all similarity values computed on the 2Dim or 3Dim parameter grids (for the grid search GS2D or GS3D) and from all goodness-of-fit values obtained by 15 independent runs of the optimization algorithms (NMA2D, PSO2D, CMAES2D, BO2D in 2Dim and NMA3D, PSO3D, CMAES3D, BO3D in 3Dim). In the violin plots of one-parameter distributions for the 2Dim (A, B) and 3Dim (D–F) cases, the methods are indicated on the horizontal axes along with the optimized parameter values on the vertical axes. The plots (C1–C5) show the distributions of the optimal model parameters in the 2Dim space, where the relative frequency of the optimal parameter values is depicted by color and the respective methods are indicated in the plots. The figure was created with MATLAB R2018a (www.mathworks.com).