Figure 5
From: Two-stage variable-fidelity modeling of antennas with domain confinement
Conceptual illustration of the training data sampling process: (i) samples distributed according to Latin Hypercube Sampling (LHS)81 (top), (ii) mapping onto the Cartesian product of the performance figure space F and [−1,1]n–N using function h1 (middle), (iii) black circles mark data samples projected onto the confined domain XS using function h2 (bottom); gray circles indicate the samples distributed over the image of design space F prior to orthogonal extension.
