Log10 (All unique sites mean squared displacements @ 600K - Ų) - log_msd_all_600

ARFS Top features

ARFS selected descriptors

---

Correlation analysis

Distance correlation

Dependency graphs

Feature learnability

---

Model performance

5-Fold CV Metrics overview

RF - MATMINER

	train_rmse	test_rmse	train_errors	test_errors	train_r2	test_r2
mean	0.04056	0.10898	0.02936	0.07886	0.98396	0.88382
min	0.04	0.1048	0.0291	0.078	0.9834	0.8696
max	0.0411	0.1143	0.0296	0.0803	0.9845	0.895
std	0.000397995	0.00360411	0.000215407	0.000776144	0.000361109	0.00866912

RF - MATMINER+LOBSTER

	train_rmse	test_rmse	train_errors	test_errors	train_r2	test_r2
mean	0.03982	0.10732	0.02878	0.07774	0.98452	0.88724
min	0.0393	0.1018	0.0286	0.0761	0.9842	0.8724
max	0.0401	0.113	0.0291	0.0801	0.985	0.8989
std	0.000292575	0.00462662	0.000193907	0.00145959	0.000263818	0.00959679

MODNet - MATMINER

	train_rmse	test_rmse	train_errors	test_errors	train_r2	test_r2
mean	0.04722	0.0906	0.0307	0.06218	0.978	0.91972
min	0.0386	0.0884	0.0244	0.0602	0.9729	0.912
max	0.0529	0.093	0.0348	0.0633	0.9855	0.9265
std	0.00474148	0.00195038	0.00354514	0.00117881	0.00417852	0.00460191

MODNet - MATMINER+LOBSTER

	train_rmse	test_rmse	train_errors	test_errors	train_r2	test_r2
mean	0.03602	0.08594	0.02336	0.05874	0.98732	0.92784
min	0.0348	0.0823	0.0223	0.0568	0.9859	0.9231
max	0.0378	0.091	0.0249	0.0609	0.9882	0.9324
std	0.00101863	0.00283591	0.00095205	0.0014207	0.000767854	0.00374785

Corrected resampled t-test on 10-fold CV

Summary

	t_stat	p_value	significance_stars	d_av	rel_improvement	percent_folds_improved
RF	2.50417	0.0168154	*	0.356097	1.47348	90
MODNet	1.65115	0.0665515		0.551488	3.04386	80

---

Model Explainer

PFI

SHAP

---

SISSO Models

Rung 1

1D descriptor

\[\begin{align*} & log\_msd\_all\_600 = -1.108 \\ & -0.199\left(\ln{ G4_{0.005\_1.0\_1.0} }\right) \end{align*}\]

2D descriptor

\[\begin{align*} & log\_msd\_all\_600 = -1.257 \\ & -0.185\left(\ln{ G4_{0.005\_1.0\_1.0} }\right) \\ & + 0.007\left(\frac{ site_{ElectronAffinity} }{ site_{Column} } \right) \end{align*}\]

Rung 2

1D descriptor

\[\begin{align*} & log\_msd\_all\_600 = -1.111 \\ & -0.231\left(\left(\sqrt[3]{ G4_{0.005\_1.0\_1.0} }\right) - \left(\frac{1.0}{ site_{Column} }\right)\right) \end{align*}\]

2D descriptor

\[\begin{align*} & log\_msd\_all\_600 = -1.456 \\ & + 0.005\left(\left(\frac{ site_{ElectronAffinity} }{ site_{Column} } \right) + \left(site_{ElectronAffinity} BOOP_{Q\_l\_2}\right)\right) \\ & + 0.203\left(\left(\sqrt{ local_{difference\_in\_Electronegativity} }\right) - \left(\ln{ G4_{0.005\_1.0\_1.0} }\right)\right) \end{align*}\]

---

Misc

ARFS n-iter convergence checks

MAE/ fold from 10-fold CV

Alternative visual summary of input data for t-test