Fig. 4: The optimization algorithm.
From: Identification of high-dielectric constant compounds from statistical design
The value \({y}_{t}^{{\rm{max}}}\) represents the currently available highest value of ϵ among all materials in the training-data. μ and σ represent the mean and standard deviation of the ANN-ensemble predicted distribution of ϵ for a material (blue dot) in the search-space. Within the predicted distribution, which is assumed as a normalized Gaussian function here, the region above \({y}_{t}^{{\rm{max}}}\) represents the region of improvement—as shown in green. If the ab initio DFPT calculation determines that the material's ϵ value exists within the green-shaded region, it will be considered as an improvement over the current best value \({y}_{t}^{{\rm{max}}}\) in training-data.
