logml.models.registry.linear

Classes

`BaseLinearClassifierModel`([params, logger])	Base class for linear classification models without inner CV.
`BaseLinearModel`([params, logger])	Base class for linear models with/without CV.
`BaseSMLinearModel`([params, logger])	Wrapper for statsmodels.***.
`ElasticNetModel`([params, logger])	Wrapper for sklearn.linear_model.ElasticNet.
`LassoLarsAICModel`([params, logger])	Wrapper for sklearn.linear_model.LassoLarsIC with AIC criterion.
`LassoLarsBICModel`([params, logger])	Wrapper for sklearn.linear_model.LassoLarsIC with BIC criterion.
`LassoModel`([params, logger])	Wrapper for sklearn.linear_model.Lasso.
`LogisticRegressionModel`([params, logger])	Wrapper for sklearn.linear_model.LogisticRegression.
`RidgeModel`([params, logger])	Wrapper for sklearn.linear_model.Ridge.

class logml.models.registry.linear.BaseLinearModel(params: Optional[dict] = None, logger=None)

Bases: logml.models.base.BaseModel

Base class for linear models with/without CV. TASK and TAGS are set.

TASK = 'regression'

TAGS = ['linear', 'regularization']

class logml.models.registry.linear.BaseLinearClassifierModel(params: Optional[dict] = None, logger=None)

Bases: logml.models.base.BaseModel

Base class for linear classification models without inner CV. TASK and TAGS are set.

TASK = 'classification'

TAGS = ['linear', 'regularization']

class logml.models.registry.linear.LogisticRegressionModel(params: Optional[dict] = None, logger=None)

Bases: logml.models.registry.linear.BaseLinearClassifierModel, logml.model_search.shap.ShapExplainable

Wrapper for sklearn.linear_model.LogisticRegression.

F_MODEL: alias of sklearn.linear_model._logistic.LogisticRegression

DEFAULT_PARAMS = {'n_jobs': -1}

PARAMS_SPACE = {'C': [0.1, 1, 10], 'fit_intercept': [True, False], 'l1_ratio': {'distribution': 'uniform', 'params': [0.0, 1.0]}, 'max_iter': [3000], 'multi_class': ['auto'], 'n_jobs': [2], 'penalty': ['elasticnet'], 'solver': ['saga'], 'tol': [0.0001]}

get_shap_init_params(ctx: Optional[logml.model_search.shap.ShapExplainerContext] = None) → dict: Parameters for shap explainer

class logml.models.registry.linear.LassoModel(params: Optional[dict] = None, logger=None)

Bases: logml.models.registry.linear.BaseLinearModel, logml.model_search.shap.ShapExplainable

Wrapper for sklearn.linear_model.Lasso.

F_MODEL: alias of sklearn.linear_model._coordinate_descent.Lasso

DEFAULT_PARAMS = {'max_iter': 3000, 'normalize': True, 'random_state': None}

PARAMS_SPACE = {'alpha': {'distribution': 'loguniform', 'params': [-12, 2]}, 'normalize': [True]}

get_shap_init_params(ctx: Optional[logml.model_search.shap.ShapExplainerContext] = None) → dict: Parameters for shap explainer

class logml.models.registry.linear.ElasticNetModel(params: Optional[dict] = None, logger=None)

Bases: logml.models.registry.linear.BaseLinearModel, logml.model_search.shap.ShapExplainable

Wrapper for sklearn.linear_model.ElasticNet.

F_MODEL: alias of sklearn.linear_model._coordinate_descent.ElasticNet

DEFAULT_PARAMS = {}

PARAMS_SPACE = {'alpha': [1.0, {'distribution': 'loguniform', 'params': [-5, 2]}], 'l1_ratio': {'distribution': 'normal', 'params': [0.5, 0.1]}, 'max_iter': [1000, 500, 1500], 'selection': ['cyclic', 'random']}

get_shap_init_params(ctx: Optional[logml.model_search.shap.ShapExplainerContext] = None) → dict: Parameters for shap explainer

class logml.models.registry.linear.RidgeModel(params: Optional[dict] = None, logger=None)

Bases: logml.models.registry.linear.BaseLinearModel, logml.model_search.shap.ShapExplainable

Wrapper for sklearn.linear_model.Ridge.

F_MODEL: alias of sklearn.linear_model._ridge.Ridge

DEFAULT_PARAMS = {'alpha': 1.0, 'max_iter': 200, 'random_state': None}

PARAMS_SPACE = {'alpha': {'distribution': 'loguniform', 'params': [-12, 2]}, 'fit_intercept': [True, False], 'max_iter': [3000], 'normalize': [False, True]}

get_shap_init_params(ctx: Optional[logml.model_search.shap.ShapExplainerContext] = None) → dict: Parameters for shap explainer

class logml.models.registry.linear.LassoLarsAICModel(params: Optional[dict] = None, logger=None)

Bases: logml.models.registry.linear.BaseLinearModel, logml.model_search.shap.ShapExplainable

Wrapper for sklearn.linear_model.LassoLarsIC with AIC criterion.

F_MODEL = functools.partial(<class 'sklearn.linear_model._least_angle.LassoLarsIC'>, criterion='aic')

DEFAULT_PARAMS = {'max_iter': 200}

PARAMS_SPACE = {'fit_intercept': [True, False], 'max_iter': [3000], 'normalize': [True, False], 'positive': [True, False]}

get_shap_init_params(ctx: Optional[logml.model_search.shap.ShapExplainerContext] = None) → dict: Parameters for shap explainer

class logml.models.registry.linear.LassoLarsBICModel(params: Optional[dict] = None, logger=None)

Bases: logml.models.registry.linear.BaseLinearModel, logml.model_search.shap.ShapExplainable

Wrapper for sklearn.linear_model.LassoLarsIC with BIC criterion.

F_MODEL = functools.partial(<class 'sklearn.linear_model._least_angle.LassoLarsIC'>, criterion='bic')

DEFAULT_PARAMS = {'max_iter': 200}

PARAMS_SPACE = {'fit_intercept': [True, False], 'max_iter': [300], 'normalize': [True, False], 'positive': [True, False]}

get_shap_init_params(ctx: Optional[logml.model_search.shap.ShapExplainerContext] = None) → dict: Parameters for shap explainer

class logml.models.registry.linear.BaseSMLinearModel(params: Optional[dict] = None, logger=None)

Bases: logml.models.base.BaseModel

Wrapper for statsmodels.***.

TAGS = ['linear', 'cv', 'sm']

FE_MODEL_ATTRIBUTE = 'coef_'

fit_fold_model(model: statsmodels.base.model.Model): Fit internal model

fit(dataset: logml.data.datasets.cv_dataset.ModelingDataset, fit_params: Optional[Dict] = None, train_final_model=False): For each CV fold fits a model.