Log10 (Max of mean Squared displacements @ 600K - Ų) - log_msd_max_600#

ARFS Top features#

ARFS selected descriptors#

Relevant Descriptors


Correlation analysis#

Distance correlation#

Distance correlation heatmap

Dependency graphs#

Dependency graph

Feature learnability#

Feature learnability


Model performance#

5-Fold CV Metrics overview#

RF - MATMINER

train_rmse

test_rmse

train_errors

test_errors

train_r2

test_r2

mean

0.04376

0.1195

0.0314

0.08616

0.98168

0.86228

min

0.0431

0.1085

0.0309

0.0823

0.9809

0.8429

max

0.0445

0.1309

0.0317

0.0909

0.9822

0.891

std

0.000488262

0.00763702

0.000275681

0.00300905

0.000526878

0.0157951

RF - MATMINER+LOBSTER

train_rmse

test_rmse

train_errors

test_errors

train_r2

test_r2

mean

0.04386

0.1193

0.0315

0.08612

0.98158

0.86272

min

0.0431

0.1092

0.0311

0.0833

0.9805

0.8423

max

0.0449

0.1311

0.032

0.0909

0.9824

0.8895

std

0.000615142

0.00777509

0.00034641

0.0027542

0.000643117

0.0157036

MODNet - MATMINER

train_rmse

test_rmse

train_errors

test_errors

train_r2

test_r2

mean

0.04008

0.10534

0.0266

0.0727

0.98444

0.89246

min

0.0348

0.0934

0.0231

0.069

0.9812

0.8595

max

0.0441

0.1184

0.0301

0.0781

0.9884

0.9127

std

0.00428084

0.00951264

0.00282984

0.00329059

0.00321907

0.0197307

MODNet - MATMINER+LOBSTER

train_rmse

test_rmse

train_errors

test_errors

train_r2

test_r2

mean

0.0451

0.1049

0.0292

0.07308

0.97944

0.89366

min

0.035

0.0937

0.0235

0.0695

0.9603

0.8717

max

0.0641

0.1168

0.0407

0.0764

0.9882

0.9109

std

0.0105896

0.00877428

0.00652595

0.00251984

0.0102763

0.0168806

Corrected resampled t-test on 10-fold CV#

Summary

t_stat

p_value

significance_stars

d_av

rel_improvement

percent_folds_improved

RF

-0.685324

0.744798

-0.0684453

-0.795431

50

MODNet

-1.33809

0.893154

-0.218705

-3.15709

10

RF t-test

MODNet t-test


Model Explainer#

PFI#

RF pfi MODNet pfi

SHAP#

RF shap MODNet Shap


Misc#

ARFS n-iter convergence checks#

Convergence

MAE/ fold from 10-fold CV#

Alternative visual summary of input data for t-test

RF per fold MAEs

MODNet per fold MAEs