Different base obj choices for the :class:`Model<BPt.Model>` are shown below The exact str indicator, as passed to the obj param is represented by the sub-heading (within "") The avaliable models are further broken down by which can workwith different problem_types. Additionally, a link to the original models documentation as well as the implemented parameter distributions are shown.
Base Class Documenation: :class:`sklearn.tree.DecisionTreeClassifier`
Param Distributions
"default"
defaults only"dt classifier dist"
max_depth: ng.p.Scalar(lower=1, upper=30).set_integer_casting() min_samples_split: ng.p.Scalar(lower=2, upper=50).set_integer_casting() class_weight: ng.p.TransitionChoice([None, 'balanced'])
Base Class Documenation: :class:`sklearn.linear_model.LogisticRegression`
Param Distributions
"base elastic"
max_iter: 1000 multi_class: 'auto' penalty: 'elasticnet' class_weight: None solver: 'saga' l1_ratio: .5"elastic classifier"
max_iter: 1000 multi_class: 'auto' penalty: 'elasticnet' class_weight: ng.p.TransitionChoice([None, 'balanced']) solver: 'saga' l1_ratio: ng.p.Scalar(lower=.01, upper=1) C: ng.p.Log(lower=1e-5, upper=1e5)"elastic clf v2"
max_iter: 1000 multi_class: 'auto' penalty: 'elasticnet' class_weight: ng.p.TransitionChoice([None, 'balanced']) solver: 'saga' l1_ratio: ng.p.Scalar(lower=.01, upper=1) C: ng.p.Log(lower=1e-2, upper=1e5)"elastic classifier extra"
max_iter: ng.p.Scalar(lower=1000, upper=10000).set_integer_casting() multi_class: 'auto' penalty: 'elasticnet' class_weight: ng.p.TransitionChoice([None, 'balanced']) solver: 'saga' l1_ratio: ng.p.Scalar(lower=.01, upper=1) C: ng.p.Log(lower=1e-5, upper=1e5) tol: ng.p.Log(lower=1e-6, upper=.01)
Base Class Documenation: :class:`sklearn.ensemble.ExtraTreesClassifier`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`sklearn.naive_bayes.GaussianNB`
Param Distributions
"base gnb"
var_smoothing: 1e-9
Base Class Documenation: :class:`sklearn.ensemble.GradientBoostingClassifier`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`sklearn.gaussian_process.GaussianProcessClassifier`
Param Distributions
"base gp classifier"
n_restarts_optimizer: 5
Base Class Documenation: :class:`sklearn.ensemble.gradient_boosting.HistGradientBoostingClassifier`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`sklearn.neighbors.KNeighborsClassifier`
Param Distributions
"base knn"
n_neighbors: 5"knn dist"
weights: ng.p.TransitionChoice(['uniform', 'distance']) n_neighbors: ng.p.Scalar(lower=2, upper=25).set_integer_casting()
Base Class Documenation: :class:`sklearn.linear_model.LogisticRegression`
Param Distributions
"base lasso"
max_iter: 1000 multi_class: 'auto' penalty: 'l1' class_weight: None solver: 'liblinear'"lasso C"
max_iter: 1000 multi_class: 'auto' penalty: 'l1' class_weight: ng.p.TransitionChoice([None, 'balanced']) solver: 'liblinear' C: ng.p.Log(lower=1e-5, upper=1e3)"lasso C extra"
max_iter: ng.p.Scalar(lower=1000, upper=10000).set_integer_casting() multi_class: 'auto' penalty: 'l1' class_weight: ng.p.TransitionChoice([None, 'balanced']) solver: 'liblinear' C: ng.p.Log(lower=1e-5, upper=1e3) tol: ng.p.Log(lower=1e-6, upper=.01)
Base Class Documenation: :class:`lightgbm.LGBMClassifier`
Param Distributions
"base lgbm"
silent: True"lgbm classifier dist1"
silent: True boosting_type: ng.p.TransitionChoice(['gbdt', 'dart', 'goss']) n_estimators: ng.p.Scalar(init=100, lower=3, upper=500).set_integer_casting() num_leaves: ng.p.Scalar(init=20, lower=6, upper=80).set_integer_casting() min_child_samples: ng.p.Scalar(lower=10, upper=500).set_integer_casting() min_child_weight: ng.p.Log(lower=1e-5, upper=1e4) subsample: ng.p.Scalar(lower=.3, upper=.95) colsample_bytree: ng.p.Scalar(lower=.3, upper=.95) reg_alpha: ng.p.TransitionChoice([0, ng.p.Log(lower=1e-5, upper=1)]) reg_lambda: ng.p.TransitionChoice([0, ng.p.Log(lower=1e-5, upper=1)]) class_weight: ng.p.TransitionChoice([None, 'balanced'])"lgbm classifier dist2"
silent: True lambda_l2: 0.001 boosting_type: ng.p.TransitionChoice(['gbdt', 'dart']) min_child_samples: ng.p.TransitionChoice([1, 5, 7, 10, 15, 20, 35, 50, 100, 200, 500, 1000]) num_leaves: ng.p.TransitionChoice([2, 4, 7, 10, 15, 20, 25, 30, 35, 40, 50, 65, 80, 100, 125, 150, 200, 250]) colsample_bytree: ng.p.TransitionChoice([0.7, 0.9, 1.0]) subsample: ng.p.Scalar(lower=.3, upper=1) learning_rate: ng.p.TransitionChoice([0.01, 0.05, 0.1]) n_estimators: ng.p.TransitionChoice([5, 20, 35, 50, 75, 100, 150, 200, 350, 500, 750, 1000]) class_weight: ng.p.TransitionChoice([None, 'balanced'])
Base Class Documenation: :class:`sklearn.svm.LinearSVC`
Param Distributions
"base linear svc"
max_iter: 1000"linear svc dist"
max_iter: 1000 C: ng.p.Log(lower=1e-4, upper=1e4) class_weight: ng.p.TransitionChoice([None, 'balanced'])
Base Class Documenation: :class:`sklearn.linear_model.LogisticRegression`
Param Distributions
"base logistic"
max_iter: 1000 multi_class: 'auto' penalty: 'none' class_weight: None solver: 'lbfgs'
Base Class Documenation: :class:`BPt.extensions.MLP.MLPClassifier_Wrapper`
Param Distributions
"default"
defaults only"mlp dist 3 layer"
hidden_layer_sizes: ng.p.Array(init=(100, 100, 100)).set_mutation(sigma=50).set_bounds(lower=1, upper=300).set_integer_casting() activation: ng.p.TransitionChoice(['identity', 'logistic', 'tanh', 'relu']) alpha: ng.p.Log(lower=1e-5, upper=1e2) batch_size: ng.p.TransitionChoice(['auto', ng.p.Scalar(init=200, lower=50, upper=400).set_integer_casting()]) learning_rate: ng.p.TransitionChoice(['constant', 'invscaling', 'adaptive']) learning_rate_init: ng.p.Log(lower=1e-5, upper=1e2) max_iter: ng.p.Scalar(init=200, lower=100, upper=1000).set_integer_casting() beta_1: ng.p.Scalar(init=.9, lower=.1, upper=.99) beta_2: ng.p.Scalar(init=.999, lower=.1, upper=.9999)"mlp dist es 3 layer"
hidden_layer_sizes: ng.p.Scalar(init=100, lower=2, upper=300).set_integer_casting() activation: ng.p.TransitionChoice(['identity', 'logistic', 'tanh', 'relu']) alpha: ng.p.Log(lower=1e-5, upper=1e2) batch_size: ng.p.TransitionChoice(['auto', ng.p.Scalar(init=200, lower=50, upper=400).set_integer_casting()]) learning_rate: ng.p.TransitionChoice(['constant', 'invscaling', 'adaptive']) learning_rate_init: ng.p.Log(lower=1e-5, upper=1e2) max_iter: ng.p.Scalar(init=200, lower=100, upper=1000).set_integer_casting() beta_1: ng.p.Scalar(init=.9, lower=.1, upper=.99) beta_2: ng.p.Scalar(init=.999, lower=.1, upper=.9999) early_stopping: True n_iter_no_change: ng.p.Scalar(lower=5, upper=50)"mlp dist 2 layer"
hidden_layer_sizes: ng.p.Array(init=(100, 100)).set_mutation(sigma=50).set_bounds(lower=1, upper=300).set_integer_casting() activation: ng.p.TransitionChoice(['identity', 'logistic', 'tanh', 'relu']) alpha: ng.p.Log(lower=1e-5, upper=1e2) batch_size: ng.p.TransitionChoice(['auto', ng.p.Scalar(init=200, lower=50, upper=400).set_integer_casting()]) learning_rate: ng.p.TransitionChoice(['constant', 'invscaling', 'adaptive']) learning_rate_init: ng.p.Log(lower=1e-5, upper=1e2) max_iter: ng.p.Scalar(init=200, lower=100, upper=1000).set_integer_casting() beta_1: ng.p.Scalar(init=.9, lower=.1, upper=.99) beta_2: ng.p.Scalar(init=.999, lower=.1, upper=.9999)"mlp dist es 2 layer"
hidden_layer_sizes: ng.p.Scalar(init=100, lower=2, upper=300).set_integer_casting() activation: ng.p.TransitionChoice(['identity', 'logistic', 'tanh', 'relu']) alpha: ng.p.Log(lower=1e-5, upper=1e2) batch_size: ng.p.TransitionChoice(['auto', ng.p.Scalar(init=200, lower=50, upper=400).set_integer_casting()]) learning_rate: ng.p.TransitionChoice(['constant', 'invscaling', 'adaptive']) learning_rate_init: ng.p.Log(lower=1e-5, upper=1e2) max_iter: ng.p.Scalar(init=200, lower=100, upper=1000).set_integer_casting() beta_1: ng.p.Scalar(init=.9, lower=.1, upper=.99) beta_2: ng.p.Scalar(init=.999, lower=.1, upper=.9999) early_stopping: True n_iter_no_change: ng.p.Scalar(lower=5, upper=50)"mlp dist 1 layer"
hidden_layer_sizes: ng.p.Scalar(init=100, lower=2, upper=300).set_integer_casting() activation: ng.p.TransitionChoice(['identity', 'logistic', 'tanh', 'relu']) alpha: ng.p.Log(lower=1e-5, upper=1e2) batch_size: ng.p.TransitionChoice(['auto', ng.p.Scalar(init=200, lower=50, upper=400).set_integer_casting()]) learning_rate: ng.p.TransitionChoice(['constant', 'invscaling', 'adaptive']) learning_rate_init: ng.p.Log(lower=1e-5, upper=1e2) max_iter: ng.p.Scalar(init=200, lower=100, upper=1000).set_integer_casting() beta_1: ng.p.Scalar(init=.9, lower=.1, upper=.99) beta_2: ng.p.Scalar(init=.999, lower=.1, upper=.9999)"mlp dist es 1 layer"
hidden_layer_sizes: ng.p.Scalar(init=100, lower=2, upper=300).set_integer_casting() activation: ng.p.TransitionChoice(['identity', 'logistic', 'tanh', 'relu']) alpha: ng.p.Log(lower=1e-5, upper=1e2) batch_size: ng.p.TransitionChoice(['auto', ng.p.Scalar(init=200, lower=50, upper=400).set_integer_casting()]) learning_rate: ng.p.TransitionChoice(['constant', 'invscaling', 'adaptive']) learning_rate_init: ng.p.Log(lower=1e-5, upper=1e2) max_iter: ng.p.Scalar(init=200, lower=100, upper=1000).set_integer_casting() beta_1: ng.p.Scalar(init=.9, lower=.1, upper=.99) beta_2: ng.p.Scalar(init=.999, lower=.1, upper=.9999) early_stopping: True n_iter_no_change: ng.p.Scalar(lower=5, upper=50)
Base Class Documenation: :class:`sklearn.linear_model.PassiveAggressiveClassifier`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`sklearn.ensemble.RandomForestClassifier`
Param Distributions
"base rf regressor"
n_estimators: 100"rf classifier dist"
n_estimators: ng.p.Scalar(init=100, lower=3, upper=500).set_integer_casting() max_depth: ng.p.TransitionChoice([None, ng.p.Scalar(init=25, lower=2, upper=200).set_integer_casting()]) max_features: ng.p.Scalar(lower=.1, upper=1.0) min_samples_split: ng.p.Scalar(lower=.1, upper=1.0) bootstrap: True class_weight: ng.p.TransitionChoice([None, 'balanced'])
Base Class Documenation: :class:`sklearn.linear_model.LogisticRegression`
Param Distributions
"base ridge"
max_iter: 1000 penalty: 'l2' solver: 'saga'"ridge C"
max_iter: 1000 solver: 'saga' C: ng.p.Log(lower=1e-5, upper=1e3) class_weight: ng.p.TransitionChoice([None, 'balanced'])"ridge C extra"
max_iter: ng.p.Scalar(lower=1000, upper=10000).set_integer_casting() solver: 'saga' C: ng.p.Log(lower=1e-5, upper=1e3) class_weight: ng.p.TransitionChoice([None, 'balanced']) tol: ng.p.Log(lower=1e-6, upper=.01)
Base Class Documenation: :class:`sklearn.linear_model.SGDClassifier`
Param Distributions
"base sgd"
loss: 'hinge'"sgd classifier"
loss: ng.p.TransitionChoice(['hinge', 'log', 'modified_huber', 'squared_hinge', 'perceptron']) penalty: ng.p.TransitionChoice(['l2', 'l1', 'elasticnet']) alpha: ng.p.Log(lower=1e-5, upper=1e2) l1_ratio: ng.p.Scalar(lower=0, upper=1) max_iter: 1000 learning_rate: ng.p.TransitionChoice(['optimal', 'invscaling', 'adaptive', 'constant']) eta0: ng.p.Log(lower=1e-6, upper=1e3) power_t: ng.p.Scalar(lower=.1, upper=.9) early_stopping: ng.p.TransitionChoice([False, True]) validation_fraction: ng.p.Scalar(lower=.05, upper=.5) n_iter_no_change: ng.p.TransitionChoice(np.arange(2, 20)) class_weight: ng.p.TransitionChoice([None, 'balanced'])
Base Class Documenation: :class:`sklearn.svm.SVC`
Param Distributions
"base svm classifier"
kernel: 'rbf' gamma: 'scale' probability: True"svm classifier dist"
kernel: 'rbf' gamma: ng.p.Log(lower=1e-6, upper=1) C: ng.p.Log(lower=1e-4, upper=1e4) probability: True class_weight: ng.p.TransitionChoice([None, 'balanced'])
Base Class Documenation: :class:`xgboost.XGBClassifier`
Param Distributions
"base xgb classifier"
verbosity: 0 objective: 'binary:logistic'"xgb classifier dist1"
verbosity: 0 objective: 'binary:logistic' n_estimators: ng.p.Scalar(init=100, lower=3, upper=500).set_integer_casting() min_child_weight: ng.p.Log(lower=1e-5, upper=1e4) subsample: ng.p.Scalar(lower=.3, upper=.95) colsample_bytree: ng.p.Scalar(lower=.3, upper=.95) reg_alpha: ng.p.TransitionChoice([0, ng.p.Log(lower=1e-5, upper=1)]) reg_lambda: ng.p.TransitionChoice([0, ng.p.Log(lower=1e-5, upper=1)])"xgb classifier dist2"
verbosity: 0 objective: 'binary:logistic' max_depth: ng.p.TransitionChoice([None, ng.p.Scalar(init=25, lower=2, upper=200).set_integer_casting()]) learning_rate: ng.p.Scalar(lower=.01, upper=.5) n_estimators: ng.p.Scalar(lower=3, upper=500).set_integer_casting() min_child_weight: ng.p.TransitionChoice([1, 5, 10, 50]) subsample: ng.p.Scalar(lower=.5, upper=1) colsample_bytree: ng.p.Scalar(lower=.4, upper=.95)"xgb classifier dist3"
verbosity: 0 objective: 'binary:logistic' learning_rare: ng.p.Scalar(lower=.005, upper=.3) min_child_weight: ng.p.Scalar(lower=.5, upper=10) max_depth: ng.p.TransitionChoice(np.arange(3, 10)) subsample: ng.p.Scalar(lower=.5, upper=1) colsample_bytree: ng.p.Scalar(lower=.5, upper=1) reg_alpha: ng.p.Log(lower=.00001, upper=1)
Base Class Documenation: :class:`sklearn.linear_model.ARDRegression`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`sklearn.linear_model.BayesianRidge`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`sklearn.tree.DecisionTreeRegressor`
Param Distributions
"default"
defaults only"dt dist"
max_depth: ng.p.Scalar(lower=1, upper=30).set_integer_casting() min_samples_split: ng.p.Scalar(lower=2, upper=50).set_integer_casting()
Base Class Documenation: :class:`sklearn.linear_model.ElasticNet`
Param Distributions
"base elastic net"
max_iter: 1000"elastic regression"
max_iter: 1000 alpha: ng.p.Log(lower=1e-5, upper=1e5) l1_ratio: ng.p.Scalar(lower=.01, upper=1)"elastic regression extra"
max_iter: ng.p.Scalar(lower=1000, upper=10000).set_integer_casting() alpha: ng.p.Log(lower=1e-5, upper=1e5) l1_ratio: ng.p.Scalar(lower=.01, upper=1) tol: ng.p.Log(lower=1e-6, upper=.01)
Base Class Documenation: :class:`sklearn.ensemble.ExtraTreesRegressor`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`sklearn.ensemble.GradientBoostingRegressor`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`sklearn.gaussian_process.GaussianProcessRegressor`
Param Distributions
"base gp regressor"
n_restarts_optimizer: 5 normalize_y: True
Base Class Documenation: :class:`sklearn.ensemble.gradient_boosting.HistGradientBoostingRegressor`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`sklearn.neighbors.KNeighborsRegressor`
Param Distributions
"base knn regression"
n_neighbors: 5"knn dist regression"
weights: ng.p.TransitionChoice(['uniform', 'distance']) n_neighbors: ng.p.Scalar(lower=2, upper=25).set_integer_casting()
Base Class Documenation: :class:`sklearn.linear_model.Lasso`
Param Distributions
"base lasso regressor"
max_iter: 1000"lasso regressor dist"
max_iter: 1000 alpha: ng.p.Log(lower=1e-5, upper=1e5)
Base Class Documenation: :class:`lightgbm.LGBMRegressor`
Param Distributions
"base lgbm"
silent: True"lgbm dist1"
silent: True boosting_type: ng.p.TransitionChoice(['gbdt', 'dart', 'goss']) n_estimators: ng.p.Scalar(init=100, lower=3, upper=500).set_integer_casting() num_leaves: ng.p.Scalar(init=20, lower=6, upper=80).set_integer_casting() min_child_samples: ng.p.Scalar(lower=10, upper=500).set_integer_casting() min_child_weight: ng.p.Log(lower=1e-5, upper=1e4) subsample: ng.p.Scalar(lower=.3, upper=.95) colsample_bytree: ng.p.Scalar(lower=.3, upper=.95) reg_alpha: ng.p.TransitionChoice([0, ng.p.Log(lower=1e-5, upper=1)]) reg_lambda: ng.p.TransitionChoice([0, ng.p.Log(lower=1e-5, upper=1)])"lgbm dist2"
silent: True lambda_l2: 0.001 boosting_type: ng.p.TransitionChoice(['gbdt', 'dart']) min_child_samples: ng.p.TransitionChoice([1, 5, 7, 10, 15, 20, 35, 50, 100, 200, 500, 1000]) num_leaves: ng.p.TransitionChoice([2, 4, 7, 10, 15, 20, 25, 30, 35, 40, 50, 65, 80, 100, 125, 150, 200, 250]) colsample_bytree: ng.p.TransitionChoice([0.7, 0.9, 1.0]) subsample: ng.p.Scalar(lower=.3, upper=1) learning_rate: ng.p.TransitionChoice([0.01, 0.05, 0.1]) n_estimators: ng.p.TransitionChoice([5, 20, 35, 50, 75, 100, 150, 200, 350, 500, 750, 1000])
Base Class Documenation: :class:`sklearn.linear_model.LinearRegression`
Param Distributions
"base linear"
fit_intercept: True
Base Class Documenation: :class:`sklearn.svm.LinearSVR`
Param Distributions
"base linear svr"
loss: 'epsilon_insensitive' max_iter: 1000"linear svr dist"
loss: 'epsilon_insensitive' max_iter: 1000 C: ng.p.Log(lower=1e-4, upper=1e4)
Base Class Documenation: :class:`BPt.extensions.MLP.MLPRegressor_Wrapper`
Param Distributions
"default"
defaults only"mlp dist 3 layer"
hidden_layer_sizes: ng.p.Array(init=(100, 100, 100)).set_mutation(sigma=50).set_bounds(lower=1, upper=300).set_integer_casting() activation: ng.p.TransitionChoice(['identity', 'logistic', 'tanh', 'relu']) alpha: ng.p.Log(lower=1e-5, upper=1e2) batch_size: ng.p.TransitionChoice(['auto', ng.p.Scalar(init=200, lower=50, upper=400).set_integer_casting()]) learning_rate: ng.p.TransitionChoice(['constant', 'invscaling', 'adaptive']) learning_rate_init: ng.p.Log(lower=1e-5, upper=1e2) max_iter: ng.p.Scalar(init=200, lower=100, upper=1000).set_integer_casting() beta_1: ng.p.Scalar(init=.9, lower=.1, upper=.99) beta_2: ng.p.Scalar(init=.999, lower=.1, upper=.9999)"mlp dist es 3 layer"
hidden_layer_sizes: ng.p.Scalar(init=100, lower=2, upper=300).set_integer_casting() activation: ng.p.TransitionChoice(['identity', 'logistic', 'tanh', 'relu']) alpha: ng.p.Log(lower=1e-5, upper=1e2) batch_size: ng.p.TransitionChoice(['auto', ng.p.Scalar(init=200, lower=50, upper=400).set_integer_casting()]) learning_rate: ng.p.TransitionChoice(['constant', 'invscaling', 'adaptive']) learning_rate_init: ng.p.Log(lower=1e-5, upper=1e2) max_iter: ng.p.Scalar(init=200, lower=100, upper=1000).set_integer_casting() beta_1: ng.p.Scalar(init=.9, lower=.1, upper=.99) beta_2: ng.p.Scalar(init=.999, lower=.1, upper=.9999) early_stopping: True n_iter_no_change: ng.p.Scalar(lower=5, upper=50)"mlp dist 2 layer"
hidden_layer_sizes: ng.p.Array(init=(100, 100)).set_mutation(sigma=50).set_bounds(lower=1, upper=300).set_integer_casting() activation: ng.p.TransitionChoice(['identity', 'logistic', 'tanh', 'relu']) alpha: ng.p.Log(lower=1e-5, upper=1e2) batch_size: ng.p.TransitionChoice(['auto', ng.p.Scalar(init=200, lower=50, upper=400).set_integer_casting()]) learning_rate: ng.p.TransitionChoice(['constant', 'invscaling', 'adaptive']) learning_rate_init: ng.p.Log(lower=1e-5, upper=1e2) max_iter: ng.p.Scalar(init=200, lower=100, upper=1000).set_integer_casting() beta_1: ng.p.Scalar(init=.9, lower=.1, upper=.99) beta_2: ng.p.Scalar(init=.999, lower=.1, upper=.9999)"mlp dist es 2 layer"
hidden_layer_sizes: ng.p.Scalar(init=100, lower=2, upper=300).set_integer_casting() activation: ng.p.TransitionChoice(['identity', 'logistic', 'tanh', 'relu']) alpha: ng.p.Log(lower=1e-5, upper=1e2) batch_size: ng.p.TransitionChoice(['auto', ng.p.Scalar(init=200, lower=50, upper=400).set_integer_casting()]) learning_rate: ng.p.TransitionChoice(['constant', 'invscaling', 'adaptive']) learning_rate_init: ng.p.Log(lower=1e-5, upper=1e2) max_iter: ng.p.Scalar(init=200, lower=100, upper=1000).set_integer_casting() beta_1: ng.p.Scalar(init=.9, lower=.1, upper=.99) beta_2: ng.p.Scalar(init=.999, lower=.1, upper=.9999) early_stopping: True n_iter_no_change: ng.p.Scalar(lower=5, upper=50)"mlp dist 1 layer"
hidden_layer_sizes: ng.p.Scalar(init=100, lower=2, upper=300).set_integer_casting() activation: ng.p.TransitionChoice(['identity', 'logistic', 'tanh', 'relu']) alpha: ng.p.Log(lower=1e-5, upper=1e2) batch_size: ng.p.TransitionChoice(['auto', ng.p.Scalar(init=200, lower=50, upper=400).set_integer_casting()]) learning_rate: ng.p.TransitionChoice(['constant', 'invscaling', 'adaptive']) learning_rate_init: ng.p.Log(lower=1e-5, upper=1e2) max_iter: ng.p.Scalar(init=200, lower=100, upper=1000).set_integer_casting() beta_1: ng.p.Scalar(init=.9, lower=.1, upper=.99) beta_2: ng.p.Scalar(init=.999, lower=.1, upper=.9999)"mlp dist es 1 layer"
hidden_layer_sizes: ng.p.Scalar(init=100, lower=2, upper=300).set_integer_casting() activation: ng.p.TransitionChoice(['identity', 'logistic', 'tanh', 'relu']) alpha: ng.p.Log(lower=1e-5, upper=1e2) batch_size: ng.p.TransitionChoice(['auto', ng.p.Scalar(init=200, lower=50, upper=400).set_integer_casting()]) learning_rate: ng.p.TransitionChoice(['constant', 'invscaling', 'adaptive']) learning_rate_init: ng.p.Log(lower=1e-5, upper=1e2) max_iter: ng.p.Scalar(init=200, lower=100, upper=1000).set_integer_casting() beta_1: ng.p.Scalar(init=.9, lower=.1, upper=.99) beta_2: ng.p.Scalar(init=.999, lower=.1, upper=.9999) early_stopping: True n_iter_no_change: ng.p.Scalar(lower=5, upper=50)
Base Class Documenation: :class:`sklearn.ensemble.RandomForestRegressor`
Param Distributions
"base rf"
n_estimators: 100"rf dist"
n_estimators: ng.p.Scalar(init=100, lower=3, upper=500).set_integer_casting() max_depth: ng.p.TransitionChoice([None, ng.p.Scalar(init=25, lower=2, upper=200).set_integer_casting()]) max_features: ng.p.Scalar(lower=.1, upper=1.0) min_samples_split: ng.p.Scalar(lower=.1, upper=1.0) bootstrap: True
Base Class Documenation: :class:`sklearn.linear_model.Ridge`
Param Distributions
"base ridge regressor"
max_iter: 1000 solver: 'lsqr'"ridge regressor dist"
max_iter: 1000 solver: 'lsqr' alpha: ng.p.Log(lower=1e-3, upper=1e5)
Base Class Documenation: :class:`sklearn.svm.SVR`
Param Distributions
"base svm"
kernel: 'rbf' gamma: 'scale'"svm dist"
kernel: 'rbf' gamma: ng.p.Log(lower=1e-6, upper=1) C: ng.p.Log(lower=1e-4, upper=1e4)
Base Class Documenation: :class:`sklearn.linear_model.glm.TweedieRegressor`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`xgboost.XGBRegressor`
Param Distributions
"base xgb"
verbosity: 0 objective: 'reg:squarederror'"xgb dist1"
verbosity: 0 objective: 'reg:squarederror' n_estimators: ng.p.Scalar(init=100, lower=3, upper=500).set_integer_casting() min_child_weight: ng.p.Log(lower=1e-5, upper=1e4) subsample: ng.p.Scalar(lower=.3, upper=.95) colsample_bytree: ng.p.Scalar(lower=.3, upper=.95) reg_alpha: ng.p.TransitionChoice([0, ng.p.Log(lower=1e-5, upper=1)]) reg_lambda: ng.p.TransitionChoice([0, ng.p.Log(lower=1e-5, upper=1)])"xgb dist2"
verbosity: 0 objective: 'reg:squarederror' max_depth: ng.p.TransitionChoice([None, ng.p.Scalar(init=25, lower=2, upper=200).set_integer_casting()]) learning_rate: ng.p.Scalar(lower=.01, upper=.5) n_estimators: ng.p.Scalar(lower=3, upper=500).set_integer_casting() min_child_weight: ng.p.TransitionChoice([1, 5, 10, 50]) subsample: ng.p.Scalar(lower=.5, upper=1) colsample_bytree: ng.p.Scalar(lower=.4, upper=.95)"xgb dist3"
verbosity: 0 objective: 'reg:squarederror' learning_rare: ng.p.Scalar(lower=.005, upper=.3) min_child_weight: ng.p.Scalar(lower=.5, upper=10) max_depth: ng.p.TransitionChoice(np.arange(3, 10)) subsample: ng.p.Scalar(lower=.5, upper=1) colsample_bytree: ng.p.Scalar(lower=.5, upper=1) reg_alpha: ng.p.Log(lower=.00001, upper=1)
Base Class Documenation: :class:`sklearn.tree.DecisionTreeClassifier`
Param Distributions
"default"
defaults only"dt classifier dist"
max_depth: ng.p.Scalar(lower=1, upper=30).set_integer_casting() min_samples_split: ng.p.Scalar(lower=2, upper=50).set_integer_casting() class_weight: ng.p.TransitionChoice([None, 'balanced'])
Base Class Documenation: :class:`sklearn.linear_model.LogisticRegression`
Param Distributions
"base elastic"
max_iter: 1000 multi_class: 'auto' penalty: 'elasticnet' class_weight: None solver: 'saga' l1_ratio: .5"elastic classifier"
max_iter: 1000 multi_class: 'auto' penalty: 'elasticnet' class_weight: ng.p.TransitionChoice([None, 'balanced']) solver: 'saga' l1_ratio: ng.p.Scalar(lower=.01, upper=1) C: ng.p.Log(lower=1e-5, upper=1e5)"elastic clf v2"
max_iter: 1000 multi_class: 'auto' penalty: 'elasticnet' class_weight: ng.p.TransitionChoice([None, 'balanced']) solver: 'saga' l1_ratio: ng.p.Scalar(lower=.01, upper=1) C: ng.p.Log(lower=1e-2, upper=1e5)"elastic classifier extra"
max_iter: ng.p.Scalar(lower=1000, upper=10000).set_integer_casting() multi_class: 'auto' penalty: 'elasticnet' class_weight: ng.p.TransitionChoice([None, 'balanced']) solver: 'saga' l1_ratio: ng.p.Scalar(lower=.01, upper=1) C: ng.p.Log(lower=1e-5, upper=1e5) tol: ng.p.Log(lower=1e-6, upper=.01)
Base Class Documenation: :class:`sklearn.ensemble.ExtraTreesClassifier`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`sklearn.naive_bayes.GaussianNB`
Param Distributions
"base gnb"
var_smoothing: 1e-9
Base Class Documenation: :class:`sklearn.ensemble.GradientBoostingClassifier`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`sklearn.gaussian_process.GaussianProcessClassifier`
Param Distributions
"base gp classifier"
n_restarts_optimizer: 5
Base Class Documenation: :class:`sklearn.ensemble.gradient_boosting.HistGradientBoostingClassifier`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`sklearn.neighbors.KNeighborsClassifier`
Param Distributions
"base knn"
n_neighbors: 5"knn dist"
weights: ng.p.TransitionChoice(['uniform', 'distance']) n_neighbors: ng.p.Scalar(lower=2, upper=25).set_integer_casting()
Base Class Documenation: :class:`sklearn.linear_model.LogisticRegression`
Param Distributions
"base lasso"
max_iter: 1000 multi_class: 'auto' penalty: 'l1' class_weight: None solver: 'liblinear'"lasso C"
max_iter: 1000 multi_class: 'auto' penalty: 'l1' class_weight: ng.p.TransitionChoice([None, 'balanced']) solver: 'liblinear' C: ng.p.Log(lower=1e-5, upper=1e3)"lasso C extra"
max_iter: ng.p.Scalar(lower=1000, upper=10000).set_integer_casting() multi_class: 'auto' penalty: 'l1' class_weight: ng.p.TransitionChoice([None, 'balanced']) solver: 'liblinear' C: ng.p.Log(lower=1e-5, upper=1e3) tol: ng.p.Log(lower=1e-6, upper=.01)
Base Class Documenation: :class:`lightgbm.LGBMClassifier`
Param Distributions
"base lgbm"
silent: True"lgbm classifier dist1"
silent: True boosting_type: ng.p.TransitionChoice(['gbdt', 'dart', 'goss']) n_estimators: ng.p.Scalar(init=100, lower=3, upper=500).set_integer_casting() num_leaves: ng.p.Scalar(init=20, lower=6, upper=80).set_integer_casting() min_child_samples: ng.p.Scalar(lower=10, upper=500).set_integer_casting() min_child_weight: ng.p.Log(lower=1e-5, upper=1e4) subsample: ng.p.Scalar(lower=.3, upper=.95) colsample_bytree: ng.p.Scalar(lower=.3, upper=.95) reg_alpha: ng.p.TransitionChoice([0, ng.p.Log(lower=1e-5, upper=1)]) reg_lambda: ng.p.TransitionChoice([0, ng.p.Log(lower=1e-5, upper=1)]) class_weight: ng.p.TransitionChoice([None, 'balanced'])"lgbm classifier dist2"
silent: True lambda_l2: 0.001 boosting_type: ng.p.TransitionChoice(['gbdt', 'dart']) min_child_samples: ng.p.TransitionChoice([1, 5, 7, 10, 15, 20, 35, 50, 100, 200, 500, 1000]) num_leaves: ng.p.TransitionChoice([2, 4, 7, 10, 15, 20, 25, 30, 35, 40, 50, 65, 80, 100, 125, 150, 200, 250]) colsample_bytree: ng.p.TransitionChoice([0.7, 0.9, 1.0]) subsample: ng.p.Scalar(lower=.3, upper=1) learning_rate: ng.p.TransitionChoice([0.01, 0.05, 0.1]) n_estimators: ng.p.TransitionChoice([5, 20, 35, 50, 75, 100, 150, 200, 350, 500, 750, 1000]) class_weight: ng.p.TransitionChoice([None, 'balanced'])
Base Class Documenation: :class:`sklearn.svm.LinearSVC`
Param Distributions
"base linear svc"
max_iter: 1000"linear svc dist"
max_iter: 1000 C: ng.p.Log(lower=1e-4, upper=1e4) class_weight: ng.p.TransitionChoice([None, 'balanced'])
Base Class Documenation: :class:`sklearn.linear_model.LogisticRegression`
Param Distributions
"base logistic"
max_iter: 1000 multi_class: 'auto' penalty: 'none' class_weight: None solver: 'lbfgs'
Base Class Documenation: :class:`BPt.extensions.MLP.MLPClassifier_Wrapper`
Param Distributions
"default"
defaults only"mlp dist 3 layer"
hidden_layer_sizes: ng.p.Array(init=(100, 100, 100)).set_mutation(sigma=50).set_bounds(lower=1, upper=300).set_integer_casting() activation: ng.p.TransitionChoice(['identity', 'logistic', 'tanh', 'relu']) alpha: ng.p.Log(lower=1e-5, upper=1e2) batch_size: ng.p.TransitionChoice(['auto', ng.p.Scalar(init=200, lower=50, upper=400).set_integer_casting()]) learning_rate: ng.p.TransitionChoice(['constant', 'invscaling', 'adaptive']) learning_rate_init: ng.p.Log(lower=1e-5, upper=1e2) max_iter: ng.p.Scalar(init=200, lower=100, upper=1000).set_integer_casting() beta_1: ng.p.Scalar(init=.9, lower=.1, upper=.99) beta_2: ng.p.Scalar(init=.999, lower=.1, upper=.9999)"mlp dist es 3 layer"
hidden_layer_sizes: ng.p.Scalar(init=100, lower=2, upper=300).set_integer_casting() activation: ng.p.TransitionChoice(['identity', 'logistic', 'tanh', 'relu']) alpha: ng.p.Log(lower=1e-5, upper=1e2) batch_size: ng.p.TransitionChoice(['auto', ng.p.Scalar(init=200, lower=50, upper=400).set_integer_casting()]) learning_rate: ng.p.TransitionChoice(['constant', 'invscaling', 'adaptive']) learning_rate_init: ng.p.Log(lower=1e-5, upper=1e2) max_iter: ng.p.Scalar(init=200, lower=100, upper=1000).set_integer_casting() beta_1: ng.p.Scalar(init=.9, lower=.1, upper=.99) beta_2: ng.p.Scalar(init=.999, lower=.1, upper=.9999) early_stopping: True n_iter_no_change: ng.p.Scalar(lower=5, upper=50)"mlp dist 2 layer"
hidden_layer_sizes: ng.p.Array(init=(100, 100)).set_mutation(sigma=50).set_bounds(lower=1, upper=300).set_integer_casting() activation: ng.p.TransitionChoice(['identity', 'logistic', 'tanh', 'relu']) alpha: ng.p.Log(lower=1e-5, upper=1e2) batch_size: ng.p.TransitionChoice(['auto', ng.p.Scalar(init=200, lower=50, upper=400).set_integer_casting()]) learning_rate: ng.p.TransitionChoice(['constant', 'invscaling', 'adaptive']) learning_rate_init: ng.p.Log(lower=1e-5, upper=1e2) max_iter: ng.p.Scalar(init=200, lower=100, upper=1000).set_integer_casting() beta_1: ng.p.Scalar(init=.9, lower=.1, upper=.99) beta_2: ng.p.Scalar(init=.999, lower=.1, upper=.9999)"mlp dist es 2 layer"
hidden_layer_sizes: ng.p.Scalar(init=100, lower=2, upper=300).set_integer_casting() activation: ng.p.TransitionChoice(['identity', 'logistic', 'tanh', 'relu']) alpha: ng.p.Log(lower=1e-5, upper=1e2) batch_size: ng.p.TransitionChoice(['auto', ng.p.Scalar(init=200, lower=50, upper=400).set_integer_casting()]) learning_rate: ng.p.TransitionChoice(['constant', 'invscaling', 'adaptive']) learning_rate_init: ng.p.Log(lower=1e-5, upper=1e2) max_iter: ng.p.Scalar(init=200, lower=100, upper=1000).set_integer_casting() beta_1: ng.p.Scalar(init=.9, lower=.1, upper=.99) beta_2: ng.p.Scalar(init=.999, lower=.1, upper=.9999) early_stopping: True n_iter_no_change: ng.p.Scalar(lower=5, upper=50)"mlp dist 1 layer"
hidden_layer_sizes: ng.p.Scalar(init=100, lower=2, upper=300).set_integer_casting() activation: ng.p.TransitionChoice(['identity', 'logistic', 'tanh', 'relu']) alpha: ng.p.Log(lower=1e-5, upper=1e2) batch_size: ng.p.TransitionChoice(['auto', ng.p.Scalar(init=200, lower=50, upper=400).set_integer_casting()]) learning_rate: ng.p.TransitionChoice(['constant', 'invscaling', 'adaptive']) learning_rate_init: ng.p.Log(lower=1e-5, upper=1e2) max_iter: ng.p.Scalar(init=200, lower=100, upper=1000).set_integer_casting() beta_1: ng.p.Scalar(init=.9, lower=.1, upper=.99) beta_2: ng.p.Scalar(init=.999, lower=.1, upper=.9999)"mlp dist es 1 layer"
hidden_layer_sizes: ng.p.Scalar(init=100, lower=2, upper=300).set_integer_casting() activation: ng.p.TransitionChoice(['identity', 'logistic', 'tanh', 'relu']) alpha: ng.p.Log(lower=1e-5, upper=1e2) batch_size: ng.p.TransitionChoice(['auto', ng.p.Scalar(init=200, lower=50, upper=400).set_integer_casting()]) learning_rate: ng.p.TransitionChoice(['constant', 'invscaling', 'adaptive']) learning_rate_init: ng.p.Log(lower=1e-5, upper=1e2) max_iter: ng.p.Scalar(init=200, lower=100, upper=1000).set_integer_casting() beta_1: ng.p.Scalar(init=.9, lower=.1, upper=.99) beta_2: ng.p.Scalar(init=.999, lower=.1, upper=.9999) early_stopping: True n_iter_no_change: ng.p.Scalar(lower=5, upper=50)
Base Class Documenation: :class:`sklearn.linear_model.PassiveAggressiveClassifier`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`sklearn.ensemble.RandomForestClassifier`
Param Distributions
"base rf regressor"
n_estimators: 100"rf classifier dist"
n_estimators: ng.p.Scalar(init=100, lower=3, upper=500).set_integer_casting() max_depth: ng.p.TransitionChoice([None, ng.p.Scalar(init=25, lower=2, upper=200).set_integer_casting()]) max_features: ng.p.Scalar(lower=.1, upper=1.0) min_samples_split: ng.p.Scalar(lower=.1, upper=1.0) bootstrap: True class_weight: ng.p.TransitionChoice([None, 'balanced'])
Base Class Documenation: :class:`sklearn.linear_model.LogisticRegression`
Param Distributions
"base ridge"
max_iter: 1000 penalty: 'l2' solver: 'saga'"ridge C"
max_iter: 1000 solver: 'saga' C: ng.p.Log(lower=1e-5, upper=1e3) class_weight: ng.p.TransitionChoice([None, 'balanced'])"ridge C extra"
max_iter: ng.p.Scalar(lower=1000, upper=10000).set_integer_casting() solver: 'saga' C: ng.p.Log(lower=1e-5, upper=1e3) class_weight: ng.p.TransitionChoice([None, 'balanced']) tol: ng.p.Log(lower=1e-6, upper=.01)
Base Class Documenation: :class:`sklearn.linear_model.SGDClassifier`
Param Distributions
"base sgd"
loss: 'hinge'"sgd classifier"
loss: ng.p.TransitionChoice(['hinge', 'log', 'modified_huber', 'squared_hinge', 'perceptron']) penalty: ng.p.TransitionChoice(['l2', 'l1', 'elasticnet']) alpha: ng.p.Log(lower=1e-5, upper=1e2) l1_ratio: ng.p.Scalar(lower=0, upper=1) max_iter: 1000 learning_rate: ng.p.TransitionChoice(['optimal', 'invscaling', 'adaptive', 'constant']) eta0: ng.p.Log(lower=1e-6, upper=1e3) power_t: ng.p.Scalar(lower=.1, upper=.9) early_stopping: ng.p.TransitionChoice([False, True]) validation_fraction: ng.p.Scalar(lower=.05, upper=.5) n_iter_no_change: ng.p.TransitionChoice(np.arange(2, 20)) class_weight: ng.p.TransitionChoice([None, 'balanced'])
Base Class Documenation: :class:`sklearn.svm.SVC`
Param Distributions
"base svm classifier"
kernel: 'rbf' gamma: 'scale' probability: True"svm classifier dist"
kernel: 'rbf' gamma: ng.p.Log(lower=1e-6, upper=1) C: ng.p.Log(lower=1e-4, upper=1e4) probability: True class_weight: ng.p.TransitionChoice([None, 'balanced'])
Base Class Documenation: :class:`xgboost.XGBClassifier`
Param Distributions
"base xgb classifier"
verbosity: 0 objective: 'binary:logistic'"xgb classifier dist1"
verbosity: 0 objective: 'binary:logistic' n_estimators: ng.p.Scalar(init=100, lower=3, upper=500).set_integer_casting() min_child_weight: ng.p.Log(lower=1e-5, upper=1e4) subsample: ng.p.Scalar(lower=.3, upper=.95) colsample_bytree: ng.p.Scalar(lower=.3, upper=.95) reg_alpha: ng.p.TransitionChoice([0, ng.p.Log(lower=1e-5, upper=1)]) reg_lambda: ng.p.TransitionChoice([0, ng.p.Log(lower=1e-5, upper=1)])"xgb classifier dist2"
verbosity: 0 objective: 'binary:logistic' max_depth: ng.p.TransitionChoice([None, ng.p.Scalar(init=25, lower=2, upper=200).set_integer_casting()]) learning_rate: ng.p.Scalar(lower=.01, upper=.5) n_estimators: ng.p.Scalar(lower=3, upper=500).set_integer_casting() min_child_weight: ng.p.TransitionChoice([1, 5, 10, 50]) subsample: ng.p.Scalar(lower=.5, upper=1) colsample_bytree: ng.p.Scalar(lower=.4, upper=.95)"xgb classifier dist3"
verbosity: 0 objective: 'binary:logistic' learning_rare: ng.p.Scalar(lower=.005, upper=.3) min_child_weight: ng.p.Scalar(lower=.5, upper=10) max_depth: ng.p.TransitionChoice(np.arange(3, 10)) subsample: ng.p.Scalar(lower=.5, upper=1) colsample_bytree: ng.p.Scalar(lower=.5, upper=1) reg_alpha: ng.p.Log(lower=.00001, upper=1)
Different availible choices for the scorer parameter are shown below. scorer is accepted by :class:`Problem_Spec<BPt.Problem_Spec>`, :class:`Param_Search<BPt.Param_Search>` and :class:`Feat_Importance<BPt.Feat_Importance>` The str indicator for each scorer is represented bythe sub-heading (within "") The avaliable scorers are further broken down by which can work with different problem_types. Additionally, a link to the original models documentation is shown.
Base Func Documenation: :func:`sklearn.metrics.accuracy_score`
Base Func Documenation: :func:`sklearn.metrics.roc_auc_score`
Base Func Documenation: :func:`sklearn.metrics.roc_auc_score`
Base Func Documenation: :func:`sklearn.metrics.roc_auc_score`
Base Func Documenation: :func:`sklearn.metrics.roc_auc_score`
Base Func Documenation: :func:`sklearn.metrics.roc_auc_score`
Base Func Documenation: :func:`sklearn.metrics.balanced_accuracy_score`
Base Func Documenation: :func:`sklearn.metrics.average_precision_score`
Base Func Documenation: :func:`sklearn.metrics.log_loss`
Base Func Documenation: :func:`sklearn.metrics.brier_score_loss`
Base Func Documenation: :func:`sklearn.metrics.precision_score`
Base Func Documenation: :func:`sklearn.metrics.precision_score`
Base Func Documenation: :func:`sklearn.metrics.precision_score`
Base Func Documenation: :func:`sklearn.metrics.precision_score`
Base Func Documenation: :func:`sklearn.metrics.precision_score`
Base Func Documenation: :func:`sklearn.metrics.recall_score`
Base Func Documenation: :func:`sklearn.metrics.recall_score`
Base Func Documenation: :func:`sklearn.metrics.recall_score`
Base Func Documenation: :func:`sklearn.metrics.recall_score`
Base Func Documenation: :func:`sklearn.metrics.recall_score`
Base Func Documenation: :func:`sklearn.metrics.f1_score`
Base Func Documenation: :func:`sklearn.metrics.f1_score`
Base Func Documenation: :func:`sklearn.metrics.f1_score`
Base Func Documenation: :func:`sklearn.metrics.f1_score`
Base Func Documenation: :func:`sklearn.metrics.f1_score`
Base Func Documenation: :func:`sklearn.metrics.jaccard_score`
Base Func Documenation: :func:`sklearn.metrics.jaccard_score`
Base Func Documenation: :func:`sklearn.metrics.jaccard_score`
Base Func Documenation: :func:`sklearn.metrics.jaccard_score`
Base Func Documenation: :func:`sklearn.metrics.jaccard_score`
Base Func Documenation: :func:`sklearn.metrics.hamming_loss`
Base Func Documenation: :func:`sklearn.metrics.matthews_corrcoef`
Base Func Documenation: :func:`sklearn.metrics.roc_auc_score`
Base Func Documenation: :func:`sklearn.metrics.explained_variance_score`
Base Func Documenation: :func:`sklearn.metrics.explained_variance_score`
Base Func Documenation: :func:`sklearn.metrics.r2_score`
Base Func Documenation: :func:`sklearn.metrics.max_error`
Base Func Documenation: :func:`sklearn.metrics.median_absolute_error`
Base Func Documenation: :func:`sklearn.metrics.median_absolute_error`
Base Func Documenation: :func:`sklearn.metrics.mean_absolute_error`
Base Func Documenation: :func:`sklearn.metrics.mean_absolute_error`
Base Func Documenation: :func:`sklearn.metrics.mean_squared_error`
Base Func Documenation: :func:`sklearn.metrics.mean_squared_error`
Base Func Documenation: :func:`sklearn.metrics.mean_squared_log_error`
Base Func Documenation: :func:`sklearn.metrics.mean_squared_log_error`
Base Func Documenation: :func:`sklearn.metrics.mean_squared_error`
Base Func Documenation: :func:`sklearn.metrics.mean_squared_error`
Base Func Documenation: :func:`sklearn.metrics.mean_poisson_deviance`
Base Func Documenation: :func:`sklearn.metrics.mean_poisson_deviance`
Base Func Documenation: :func:`sklearn.metrics.mean_gamma_deviance`
Base Func Documenation: :func:`sklearn.metrics.mean_gamma_deviance`
Base Func Documenation: :func:`sklearn.metrics.r2_score`
Base Func Documenation: :func:`sklearn.metrics.accuracy_score`
Base Func Documenation: :func:`sklearn.metrics.roc_auc_score`
Base Func Documenation: :func:`sklearn.metrics.roc_auc_score`
Base Func Documenation: :func:`sklearn.metrics.roc_auc_score`
Base Func Documenation: :func:`sklearn.metrics.roc_auc_score`
Base Func Documenation: :func:`sklearn.metrics.roc_auc_score`
Base Func Documenation: :func:`sklearn.metrics.balanced_accuracy_score`
Base Func Documenation: :func:`sklearn.metrics.average_precision_score`
Base Func Documenation: :func:`sklearn.metrics.log_loss`
Base Func Documenation: :func:`sklearn.metrics.brier_score_loss`
Base Func Documenation: :func:`sklearn.metrics.precision_score`
Base Func Documenation: :func:`sklearn.metrics.precision_score`
Base Func Documenation: :func:`sklearn.metrics.precision_score`
Base Func Documenation: :func:`sklearn.metrics.precision_score`
Base Func Documenation: :func:`sklearn.metrics.precision_score`
Base Func Documenation: :func:`sklearn.metrics.recall_score`
Base Func Documenation: :func:`sklearn.metrics.recall_score`
Base Func Documenation: :func:`sklearn.metrics.recall_score`
Base Func Documenation: :func:`sklearn.metrics.recall_score`
Base Func Documenation: :func:`sklearn.metrics.recall_score`
Base Func Documenation: :func:`sklearn.metrics.f1_score`
Base Func Documenation: :func:`sklearn.metrics.f1_score`
Base Func Documenation: :func:`sklearn.metrics.f1_score`
Base Func Documenation: :func:`sklearn.metrics.f1_score`
Base Func Documenation: :func:`sklearn.metrics.f1_score`
Base Func Documenation: :func:`sklearn.metrics.jaccard_score`
Base Func Documenation: :func:`sklearn.metrics.jaccard_score`
Base Func Documenation: :func:`sklearn.metrics.jaccard_score`
Base Func Documenation: :func:`sklearn.metrics.jaccard_score`
Base Func Documenation: :func:`sklearn.metrics.jaccard_score`
Base Func Documenation: :func:`sklearn.metrics.hamming_loss`
Base Func Documenation: :func:`sklearn.metrics.matthews_corrcoef`
Base Func Documenation: :func:`sklearn.metrics.roc_auc_score`
Different base obj choices for the :class:`Loader<BPt.Loader>` are shown below The exact str indicator, as passed to the obj param is represented by the sub-heading (within "") Additionally, a link to the original models documentation as well as the implemented parameter distributions are shown.
Base Class Documenation: :class:`BPt.extensions.Loaders.Identity`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`BPt.extensions.Loaders.SurfLabels`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`nilearn.input_data.nifti_labels_masker.NiftiLabelsMasker`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`BPt.extensions.Loaders.Connectivity`
Param Distributions
"default"
defaults only
Different base obj choices for the :class:`Imputer<BPt.Imputer>` are shown below The exact str indicator, as passed to the obj param is represented by the sub-heading (within "") Additionally, a link to the original models documentation as well as the implemented parameter distributions are shown. Note that if the iterative imputer is requested, base_model must also be passed.
Base Class Documenation: :class:`sklearn.impute.SimpleImputer`
Param Distributions
"mean imp"
strategy: 'mean'
Base Class Documenation: :class:`sklearn.impute.SimpleImputer`
Param Distributions
"median imp"
strategy: 'median'
Base Class Documenation: :class:`sklearn.impute.SimpleImputer`
Param Distributions
"most freq imp"
strategy: 'most_frequent'
Base Class Documenation: :class:`sklearn.impute.SimpleImputer`
Param Distributions
"constant imp"
strategy: 'constant'
Base Class Documenation: :class:`sklearn.impute.IterativeImputer`
Param Distributions
"iterative imp"
initial_strategy: 'mean' skip_complete: True
Different base obj choices for the :class:`Scaler<BPt.Scaler>` are shown below The exact str indicator, as passed to the obj param is represented by the sub-heading (within "") Additionally, a link to the original models documentation as well as the implemented parameter distributions are shown.
Base Class Documenation: :class:`sklearn.preprocessing.StandardScaler`
Param Distributions
"base standard"
with_mean: True with_std: True
Base Class Documenation: :class:`sklearn.preprocessing.MinMaxScaler`
Param Distributions
"base minmax"
feature_range: (0, 1)
Base Class Documenation: :class:`sklearn.preprocessing.MaxAbsScaler`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`sklearn.preprocessing.RobustScaler`
Param Distributions
"base robust"
quantile_range: (5, 95)"robust gs"
quantile_range: ng.p.TransitionChoice([(x, 100-x) for x in np.arange(1, 40)])
Base Class Documenation: :class:`sklearn.preprocessing.PowerTransformer`
Param Distributions
"base yeo"
method: 'yeo-johnson' standardize: True
Base Class Documenation: :class:`sklearn.preprocessing.PowerTransformer`
Param Distributions
"base boxcox"
method: 'box-cox' standardize: True
Base Class Documenation: :class:`BPt.extensions.Scalers.Winsorizer`
Param Distributions
"base winsorize"
quantile_range: (1, 99)"winsorize gs"
quantile_range: ng.p.TransitionChoice([(x, 100-x) for x in np.arange(1, 40)])
Base Class Documenation: :class:`sklearn.preprocessing.QuantileTransformer`
Param Distributions
"base quant norm"
output_distribution: 'normal'
Base Class Documenation: :class:`sklearn.preprocessing.QuantileTransformer`
Param Distributions
"base quant uniform"
output_distribution: 'uniform'
Base Class Documenation: :class:`sklearn.preprocessing.Normalizer`
Param Distributions
"default"
defaults only
Different base obj choices for the :class:`Transformer<BPt.Transformer>` are shown below The exact str indicator, as passed to the obj param is represented by the sub-heading (within "") Additionally, a link to the original models documentation as well as the implemented parameter distributions are shown.
Base Class Documenation: :class:`sklearn.decomposition.PCA`
Param Distributions
"default"
defaults only"pca var search"
n_components: ng.p.Scalar(init=.75, lower=.1, upper=.99) svd_solver: 'full'
Base Class Documenation: :class:`sklearn.decomposition.SparsePCA`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`sklearn.decomposition.MiniBatchSparsePCA`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`sklearn.decomposition.FactorAnalysis`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`sklearn.decomposition.DictionaryLearning`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`sklearn.decomposition.MiniBatchDictionaryLearning`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`sklearn.decomposition.FastICA`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`sklearn.decomposition.IncrementalPCA`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`sklearn.decomposition.KernelPCA`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`sklearn.decomposition.NMF`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`sklearn.decomposition.TruncatedSVD`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`sklearn.preprocessing.OneHotEncoder`
Param Distributions
"ohe"
sparse: False handle_unknown: 'ignore'
Base Class Documenation: :class:`category_encoders.backward_difference.BackwardDifferenceEncoder`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`category_encoders.binary.BinaryEncoder`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`category_encoders.cat_boost.CatBoostEncoder`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`category_encoders.helmert.HelmertEncoder`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`category_encoders.james_stein.JamesSteinEncoder`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`category_encoders.leave_one_out.LeaveOneOutEncoder`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`category_encoders.m_estimate.MEstimateEncoder`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`category_encoders.polynomial.PolynomialEncoder`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`category_encoders.sum_coding.SumEncoder`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`category_encoders.target_encoder.TargetEncoder`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`category_encoders.woe.WOEEncoder`
Param Distributions
"default"
defaults only
Different base obj choices for the :class:`Feat_Selector<BPt.Feat_Selector>` are shown below The exact str indicator, as passed to the obj param is represented by the sub-heading (within "") The avaliable feat selectors are further broken down by which can workwith different problem_types. Additionally, a link to the original models documentation as well as the implemented parameter distributions are shown.
Base Class Documenation: :class:`sklearn.feature_selection.RFE`
Param Distributions
"base rfe"
n_features_to_select: None"rfe num feats dist"
n_features_to_select: ng.p.Scalar(init=.5, lower=.1, upper=.99)
Base Class Documenation: :class:`BPt.extensions.Feat_Selectors.FeatureSelector`
Param Distributions
"random"
mask: 'sets as random features'"searchable"
mask: 'sets as hyperparameters'
Base Class Documenation: :class:`sklearn.feature_selection.SelectPercentile`
Param Distributions
"base univar fs classifier"
score_func: f_classif percentile: 50"univar fs classifier dist"
score_func: f_classif percentile: ng.p.Scalar(init=50, lower=1, upper=99)"univar fs classifier dist2"
score_func: f_classif percentile: ng.p.Scalar(init=75, lower=50, upper=99)
Base Class Documenation: :class:`sklearn.feature_selection.VarianceThreshold`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`sklearn.feature_selection.RFE`
Param Distributions
"base rfe"
n_features_to_select: None"rfe num feats dist"
n_features_to_select: ng.p.Scalar(init=.5, lower=.1, upper=.99)
Base Class Documenation: :class:`BPt.extensions.Feat_Selectors.FeatureSelector`
Param Distributions
"random"
mask: 'sets as random features'"searchable"
mask: 'sets as hyperparameters'
Base Class Documenation: :class:`sklearn.feature_selection.SelectPercentile`
Param Distributions
"base univar fs regression"
score_func: f_regression percentile: 50"univar fs regression dist"
score_func: f_regression percentile: ng.p.Scalar(init=50, lower=1, upper=99)"univar fs regression dist2"
score_func: f_regression percentile: ng.p.Scalar(init=75, lower=50, upper=99)
Base Class Documenation: :class:`sklearn.feature_selection.VarianceThreshold`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`sklearn.feature_selection.RFE`
Param Distributions
"base rfe"
n_features_to_select: None"rfe num feats dist"
n_features_to_select: ng.p.Scalar(init=.5, lower=.1, upper=.99)
Base Class Documenation: :class:`BPt.extensions.Feat_Selectors.FeatureSelector`
Param Distributions
"random"
mask: 'sets as random features'"searchable"
mask: 'sets as hyperparameters'
Base Class Documenation: :class:`sklearn.feature_selection.SelectPercentile`
Param Distributions
"base univar fs classifier"
score_func: f_classif percentile: 50"univar fs classifier dist"
score_func: f_classif percentile: ng.p.Scalar(init=50, lower=1, upper=99)"univar fs classifier dist2"
score_func: f_classif percentile: ng.p.Scalar(init=75, lower=50, upper=99)
Base Class Documenation: :class:`sklearn.feature_selection.VarianceThreshold`
Param Distributions
"default"
defaults only
Different base obj choices for the :class:`Ensemble<BPt.Ensemble>` are shown below The exact str indicator, as passed to the obj param is represented by the sub-heading (within "") The avaliable ensembles are further broken down by which can workwith different problem_types. Additionally, a link to the original models documentation as well as the implemented parameter distributions are shown. Also note that ensemble require a few extra params! I.e., in general, all DESlib based ensemble need needs_split = True
Base Class Documenation: :class:`sklearn.ensemble.AdaBoostClassifier`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.dcs.a_posteriori.APosteriori`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.dcs.a_priori.APriori`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`sklearn.ensemble.BaggingClassifier`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`imblearn.ensemble.BalancedBaggingClassifier`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.des.des_clustering.DESClustering`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.des.des_knn.DESKNN`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.des.probabilistic.DESKL`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.des.des_mi.DESMI`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.des.des_p.DESP`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.des.probabilistic.Exponential`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.des.knop.KNOP`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.des.knora_e.KNORAE`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.des.knora_u.KNORAU`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.dcs.lca.LCA`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.des.probabilistic.Logarithmic`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.dcs.mcb.MCB`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.des.meta_des.METADES`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.des.probabilistic.MinimumDifference`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.dcs.mla.MLA`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.dcs.ola.OLA`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.dcs.rank.Rank`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.des.probabilistic.RRC`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.static.single_best.SingleBest`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.static.stacked.StackedClassifier`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`BPt.pipeline.Ensembles.BPtStackingClassifier`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`BPt.pipeline.Ensembles.BPtVotingClassifier`
Param Distributions
"voting classifier"
voting: 'soft'
Base Class Documenation: :class:`sklearn.ensemble.AdaBoostRegressor`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`sklearn.ensemble.BaggingRegressor`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`BPt.pipeline.Ensembles.BPtStackingRegressor`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`BPt.pipeline.Ensembles.BPtVotingRegressor`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`sklearn.ensemble.AdaBoostClassifier`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.dcs.a_posteriori.APosteriori`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.dcs.a_priori.APriori`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`sklearn.ensemble.BaggingClassifier`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`imblearn.ensemble.BalancedBaggingClassifier`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.des.des_clustering.DESClustering`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.des.des_knn.DESKNN`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.des.probabilistic.DESKL`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.des.des_mi.DESMI`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.des.des_p.DESP`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.des.probabilistic.Exponential`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.des.knop.KNOP`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.des.knora_e.KNORAE`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.des.knora_u.KNORAU`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.dcs.lca.LCA`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.des.probabilistic.Logarithmic`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.dcs.mcb.MCB`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.des.meta_des.METADES`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.des.probabilistic.MinimumDifference`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.dcs.mla.MLA`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.dcs.ola.OLA`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.dcs.rank.Rank`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.des.probabilistic.RRC`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.static.single_best.SingleBest`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`deslib.static.stacked.StackedClassifier`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`BPt.pipeline.Ensembles.BPtStackingClassifier`
Param Distributions
"default"
defaults only
Base Class Documenation: :class:`BPt.pipeline.Ensembles.BPtVotingClassifier`
Param Distributions
"voting classifier"
voting: 'soft'