aim2dat.ml.cell_grid_search

Methods to fit the cell parameters of crystalline materials.

Module Contents

Classes

CellGridSearch

Class to fit the cell parameters of an initial structure to a final structure using a
class aim2dat.ml.cell_grid_search.CellGridSearch(length_scaling_factors=[0.8, 1.0, 1.2], angle_scaling_factors=[0.9, 1.0, 1.1], symprec=0.005, angle_tolerance=-1.0, hall_number=0, ffprint_r_max=10.0, ffprint_delta_bin=0.005, ffprint_sigma=0.05, ffprint_use_weights=True, ffprint_distinguish_kinds=False, target_value=0.0)[source]

Class to fit the cell parameters of an initial structure to a final structure using a brute-force grid search approach. The space group is maintained during the fitting process.

Variables:
length_scaling_factors : list

Scaling factors for the cell lengths.

angle_scaling_factors : list

Scaling factors for the cell angles.

symprec : float

Tolerance for spglib and length and angle comparison.

angle_tolerance : float

Tolerance parameter for spglib.

hall_number : int (optional)

The argument to constrain the space-group-type search only for the Hall symbol corresponding to it.

ffprint_r_max : float

Cut-off value for the maximum distance between two atoms.

ffprint_delta_bin : float (optional)

Bin size to descritize the function.

ffprint_sigma : float (optional)

Smearing parameter for the Gaussian function.

ffprint_use_weights : bool (optional)

Whether to use importance weights for the element pairs.

ffprint_distinguish_kinds : bool (optional)

Whether different kinds should be distinguished e.g. Ni0 and Ni1 would be considered as different elements if True.

target_value : float (optional)

Target value used to calculate score if a model is set via the set_model function.

Overview

Methods

fit(search_space)

Fit the initial to the final structure by varying the cell parameters.

get_optimized_structure()

Get optimized structure with the lowest score.

return_initial_score()

Return score of the initial structure.

return_search_space()

Return list of parameter sets that are varied to fit the initial to the final structure.

set_initial_structure(structure)

Set initial crystal structure.

set_model(model, function_name, single, transformer)

Set scikit-learn model to predict the target value.

set_target_structure(structure)

Set target crystal structure.
fit(search_space=None)[source]

Fit the initial to the final structure by varying the cell parameters.

Parameters:

search_space (list or None) – Defines the cell parameter variations. If set to None the parameters are obtained via the return_search_space-function.

Returns:

  • max_score (float) – Score of the best match.

  • max_params (list) – Parameters that give the best match.

get_optimized_structure()[source]

Get optimized structure with the lowest score.

Returns:

aim2dat.strct.Structure – Optimized structure.

return_initial_score()[source]

Return score of the initial structure.

Returns:

float – Score of the initial structure.

return_search_space()[source]

Return list of parameter sets that are varied to fit the initial to the final structure.

Returns:

list – List of parameter sets that are varied.

set_initial_structure(structure)[source]

Set initial crystal structure.

Parameters:

structure (aim2dat.strct.Structure) – Initial structure.

set_model(model, function_name='predict', single=False, transformer=None)[source]

Set scikit-learn model to predict the target value.

Parameters:

  • model – Object that takes structures or features as input to predicts a target value.

  • function_name (str (optional)) – Function name to retrieve the property prediction.

  • single (bool (optional)) – Whether a single structure/features or a list of structures/features is predicted at once.

  • transformer (aim2dat.ml.transformers (optional)) – Structure transformer.

set_target_structure(structure)[source]

Set target crystal structure.

Parameters:

structure (aim2dat.strct.Structure) – Target structure.