feijoa.importance package#

Submodules#

feijoa.importance.evaluator module#

Base importance evaluator class module.

class feijoa.importance.evaluator.ImportanceEvaluator#

Bases: object

Base importance evaluator.

Measures the excessive amount of hyperparameter taken per feature value. Helps to select only the most influential hyperparameters for their tuning. Does not require additional measurements.

Note

This version does not apply the study of the influence of a combination of hyperparameters.

abstract do(job)#

Calculate hyperparameters importance.

Parameters

job (Job) – Specified job.

Returns

Dict with keys params and importances

params contains names of parameters, importances contains list of importances.

Raises

AnyError – If anything bad happens.

feijoa.importance.functional_anova module#

feijoa.importance.mdi module#

MDI importance evaluator module.

class feijoa.importance.mdi.MDIEvaluator(*, n_trees=64, max_depth=64)#

Bases: feijoa.importance.evaluator.ImportanceEvaluator

Mean decrease impurity (MDI) importance evaluator.

from feijoa.importance.mdi import MDIEvaluator

job = ...
evaluator = MDIEvaluator()
imp = evaluator.do(job)

params = imp["params"]
importances = imp["importances"]
Parameters

  • n_trees (int) – number of trees in RandomForestRegressor.

  • max_depth (int) – maximum of trees depth.

do(job)#

Calculate hyperparameters importance.

Parameters

job (Job) – Specified job.

Returns

Dict with keys params and importances

params contains names of parameters, importances contains list of importances.

Raises

AnyError – If anything bad happens.

feijoa.importance.pca module#

PCA importance evaluator module.

class feijoa.importance.pca.PCAEvaluator#

Bases: feijoa.importance.evaluator.ImportanceEvaluator

Principal component analysis (PCA) importance evaluator.

from feijoa.importance.pca import PCAEvaluator

job = ...
evaluator = PCAEvaluator()
imp = evaluator.do(job)

params = imp["params"]
importances = imp["importances"]
do(job)#

Calculate hyperparameters importance.

Parameters

job (Job) – Specified job.

Returns

Dict with keys params and importances

params contains names of parameters, importances contains list of importances.

Raises

AnyError – If anything bad happens.

feijoa.importance.rsfanova_boosted module#

Module contents#