LearningLinear

Linear and generalised-linear models. Trainers in the scikit-learn Linear family, behind Learning's typed contract. Each requires a clean training set (``independent_target`` and no missing values) and stamps the model with the dataset's feature/class count. ``open LearningLinear`` brings the core types in scope too.
Imports
open LearningCore;
Table of Contents

Functions

logistic_regression

def logistic_regression (1 ds : Dataset | independent_target ds = True && has_missing ds = False) : {m : SklearnClassifier | clf_features m = ds_features ds && clf_classes m = ds_classes ds}

ridge_classifier

No predict_proba; classify_one falls back to a one-hot of predict.
def ridge_classifier (1 ds : Dataset | independent_target ds = True && has_missing ds = False) : {m : SklearnClassifier | clf_features m = ds_features ds && clf_classes m = ds_classes ds}

sgd_classifier

Default hinge loss has no predict_proba; one-hot fallback applies.
def sgd_classifier (1 ds : Dataset | independent_target ds = True && has_missing ds = False) : {m : SklearnClassifier | clf_features m = ds_features ds && clf_classes m = ds_classes ds}

perceptron

No predict_proba; one-hot fallback.
def perceptron (1 ds : Dataset | independent_target ds = True && has_missing ds = False) : {m : SklearnClassifier | clf_features m = ds_features ds && clf_classes m = ds_classes ds}

passive_aggressive_classifier

No predict_proba; one-hot fallback.
def passive_aggressive_classifier (1 ds : Dataset | independent_target ds = True && has_missing ds = False) : {m : SklearnClassifier | clf_features m = ds_features ds && clf_classes m = ds_classes ds}

linear_regression

def linear_regression (1 ds : Dataset | independent_target ds = True && has_missing ds = False) : {m : SklearnRegressor | reg_features m = ds_features ds}

ridge

def ridge (1 ds : Dataset | independent_target ds = True && has_missing ds = False) : {m : SklearnRegressor | reg_features m = ds_features ds}

ridge_alpha

def ridge_alpha (alpha : Float | alpha ≥ 0.0) (1 ds : Dataset | independent_target ds = True && has_missing ds = False) : {m : SklearnRegressor | reg_features m = ds_features ds}

lasso

def lasso (1 ds : Dataset | independent_target ds = True && has_missing ds = False) : {m : SklearnRegressor | reg_features m = ds_features ds}

lasso_alpha

def lasso_alpha (alpha : Float | alpha ≥ 0.0) (1 ds : Dataset | independent_target ds = True && has_missing ds = False) : {m : SklearnRegressor | reg_features m = ds_features ds}

elastic_net

def elastic_net (1 ds : Dataset | independent_target ds = True && has_missing ds = False) : {m : SklearnRegressor | reg_features m = ds_features ds}

bayesian_ridge

def bayesian_ridge (1 ds : Dataset | independent_target ds = True && has_missing ds = False) : {m : SklearnRegressor | reg_features m = ds_features ds}

ard_regression

def ard_regression (1 ds : Dataset | independent_target ds = True && has_missing ds = False) : {m : SklearnRegressor | reg_features m = ds_features ds}

sgd_regressor

def sgd_regressor (1 ds : Dataset | independent_target ds = True && has_missing ds = False) : {m : SklearnRegressor | reg_features m = ds_features ds}

huber_regressor

def huber_regressor (1 ds : Dataset | independent_target ds = True && has_missing ds = False) : {m : SklearnRegressor | reg_features m = ds_features ds}

passive_aggressive_regressor

def passive_aggressive_regressor (1 ds : Dataset | independent_target ds = True && has_missing ds = False) : {m : SklearnRegressor | reg_features m = ds_features ds}

theil_sen_regressor

def theil_sen_regressor (1 ds : Dataset | independent_target ds = True && has_missing ds = False) : {m : SklearnRegressor | reg_features m = ds_features ds}

ransac_regressor

def ransac_regressor (1 ds : Dataset | independent_target ds = True && has_missing ds = False) : {m : SklearnRegressor | reg_features m = ds_features ds}

orthogonal_matching_pursuit

def orthogonal_matching_pursuit (1 ds : Dataset | independent_target ds = True && has_missing ds = False) : {m : SklearnRegressor | reg_features m = ds_features ds}

lars

def lars (1 ds : Dataset | independent_target ds = True && has_missing ds = False) : {m : SklearnRegressor | reg_features m = ds_features ds}

lasso_lars

def lasso_lars (1 ds : Dataset | independent_target ds = True && has_missing ds = False) : {m : SklearnRegressor | reg_features m = ds_features ds}

poisson_regressor

Requires a non-negative target.
def poisson_regressor (1 ds : Dataset | independent_target ds = True && has_missing ds = False) : {m : SklearnRegressor | reg_features m = ds_features ds}

gamma_regressor

Requires a strictly-positive target.
def gamma_regressor (1 ds : Dataset | independent_target ds = True && has_missing ds = False) : {m : SklearnRegressor | reg_features m = ds_features ds}

tweedie_regressor

def tweedie_regressor (1 ds : Dataset | independent_target ds = True && has_missing ds = False) : {m : SklearnRegressor | reg_features m = ds_features ds}

quantile_regressor

Predicts the median by default.
def quantile_regressor (1 ds : Dataset | independent_target ds = True && has_missing ds = False) : {m : SklearnRegressor | reg_features m = ds_features ds}