Fig. 4: Comparison of methods for materials science challenges.
From: A multi-fidelity machine learning approach to high throughput materials screening
Normalised Regret vs. cost expenditure for the single fidelity EI, multi-fidelityTVR-EI, a composite of computational funnels and random search algorithms applied to the materials discovery challenges (a) Alexandria, b HOPV-15 and (c) Chen. The composite funnel displays results associated with separately provisioned funnels one for each potential budget value and the associated final regret of these funnels, it thus represents the best case scenario for the method. Median regret values are plotted from 15 optimisations with different random seeds. Shading shows the interquartile range of the runs.
