Table 1 The MAE after bias correction for ML models trained on 3 different subsets of features

From: Cognitive aging clock based on Explainable artificial intelligence

 

III

I

V

Models

train

test

train

test

train

test

LightGBM

7.5

9.1

8.8

8.9

6.9

7.6

Gradient boosting

7.7

9.1

8.4

8.6

7.0

7.8

AdaBoost

9.0

9.5

10.4

10.5

8.7

8.6

  1. Three subsets of features: III (3 tests as in ref. 28: 2 sensomotor tests on arithmetic and letters along with campimetry test), I (Stroop test), V (5 tests: 3 sensomotor tests on color memorization, arithmetic and letters and along with a Stroop and campimetry tests), cf. Section “Feature subsets”.