jarvis.ai.pkgs package

Subpackages

• jarvis.ai.pkgs.lgbm package
  • Submodules
  • jarvis.ai.pkgs.lgbm.classification module
  • jarvis.ai.pkgs.lgbm.regression module
  • Module contents
• jarvis.ai.pkgs.sklearn package
  • Submodules
  • jarvis.ai.pkgs.sklearn.classification module
  • jarvis.ai.pkgs.sklearn.hyper_params module
  • jarvis.ai.pkgs.sklearn.regression module
  • Module contents

Submodules

jarvis.ai.pkgs.utils module

Helper functions for ML applications.

jarvis.ai.pkgs.utils.binary_class_dat(X=[], Y=[], tol=0.1)

Categorize a continous dataset in 1/0 with a threshold “tol”.

TODO: replace with OneHotEncoder

jarvis.ai.pkgs.utils.get_ml_data(ml_property='formation_energy_peratom', dataset='cfid_3d', data_ranges={'bulk_modulus_kv': [0, 250], 'dfpt_piezo_max_dielectric': [0, 100], 'dfpt_piezo_max_dij': [0, 3000], 'dfpt_piezo_max_eij': [0, 10], 'ehull': [0, 1], 'electron_avg_effective_masses_300K': [0, 3], 'encut': [0, 2000], 'epsx': [0, 60], 'epsy': [0, 60], 'epsz': [0, 60], 'exfoliation_energy': [0, 1000], 'formation_energy_peratom': [-5, 5], 'hole_avg_effective_masses_300K': [0, 3], 'kpoint_length_unit': [0, 200], 'magmom_oszicar': [0, 10], 'max_ir_mode': [0, 4000], 'mbj_bandgap': [0, 10], 'mepsx': [0, 60], 'mepsy': [0, 60], 'mepsz': [0, 60], 'n-Seebeck': [-600, 10], 'n-powerfact': [0, 5000], 'optb88vdw_bandgap': [0, 10], 'p-Seebeck': [-10, 600], 'p-powerfact': [0, 5000], 'shear_modulus_gv': [0, 250], 'slme': [0, 40], 'spillage': [0, 4], 'total_energy_per_atom': [-10, 3]})

Provide arrays/pandas-dataframe as input for ML algorithms.

Args:

ml_property: target property to train

data_ranges: range for filtering data

dataset: dataset available in jarvis or other array

Returns:

X, Y , ids

jarvis.ai.pkgs.utils.regr_scores(test, pred)

Provide generic regresion scores.

Args:

pred: predicted values

test: held data for testing

Returns:

info: with metrics

Module contents

Applications of AI packages.