search

MaterSim / PyXtal_FF

Machine Learning Interatomic Potential Predictions
86 stars 23 forks source link

Low Predictability #2

Closed yanxon closed 5 years ago

yanxon commented 5 years ago 
   train_r2   train_mae   test_r2    test_mae
0   0.79307  106.110166  0.286238  351.606481

More training data!

yanxon commented 5 years ago

Change mean_squared_error to mean_absolute_error & more training data from Elastic.json, Surface.json, and Vacancy.json.

This is the time it takes to convert json files to structures: 2.36 s
This is the time it takes to generate bispectrum coefficients for 294 structures: 310.01 s
This is the time it takes running the linear regression: 0.0 s

   train_r2  train_mae   test_r2   test_mae
0  0.900277  14.575917  0.882445  16.885403