Fig. 7
From: A method for efficient Bayesian optimization of self-assembly systems from scattering data
Results from 20 rounds of optimization for a large (hypersphere of radius 9 logs) search space. Individual rounds are not delineated due to variable numbers of points returned from acquisition function minimization in different rounds. Shown are (top) the RMSD values, and (bottom) the errors in predicted parameter sets as the search progresses. Note that the best points found (very low RMSD) often correspond to comparatively distant parameter sets. This is a result of the fact that the objective can be insensitive to large displacements in some of the input dimensions but not others