Fig. 6: Examples of fitted and model-predicted composition profiles.

Composition profiles for the fcc CoCrFeMnNi high-entropy alloys: a, b, c, and d concern bounds of the composition profiles due to uncertainty of the diffusion database; e, f, g, and h concern model-predicted composition profiles due to optimization results of different algorithms, i.e., variable-selection genetic algorithm (VarSelGA), L1 norm regularization (L1 Re), L2 norm regularization (L2 Re), and Markov chain Monte Carlo (MCMC). The composition profiles of the fitted data (noted as fit) are ported from Tsai et al.⁶ for (a) and (e), Kucza et al.⁷ for (b) and (f), Dąbrowa et al.⁸ for (c) and (g), and Chen and Zhang¹¹ for (d) and (h).