Module settings

Expand description

§Settings customization

This module contains capabilities for the detailed customization of algorithm settings.

§Complete regression customization

use automl::settings::{
    Algorithm, DecisionTreeRegressorParameters, Distance, ElasticNetParameters,
    KNNAlgorithmName, KNNRegressorParameters, KNNWeightFunction, Kernel, LassoParameters,
    LinearRegressionParameters, LinearRegressionSolverName, Metric,
    RandomForestRegressorParameters, RidgeRegressionParameters, RidgeRegressionSolverName,
    SVRParameters,
 };

 let settings = automl::Settings::default_regression()
    .with_number_of_folds(3)
    .shuffle_data(true)
    .verbose(true)
    .skip(Algorithm::RandomForestRegressor)
    .sorted_by(Metric::RSquared)
    .with_linear_settings(
        LinearRegressionParameters::default().with_solver(LinearRegressionSolverName::QR),
    )
    .with_lasso_settings(
        LassoParameters::default()
            .with_alpha(10.0)
            .with_tol(1e-10)
            .with_normalize(true)
            .with_max_iter(10_000),
    )
    .with_ridge_settings(
        RidgeRegressionParameters::default()
            .with_alpha(10.0)
            .with_normalize(true)
            .with_solver(RidgeRegressionSolverName::Cholesky),
    )
    .with_elastic_net_settings(
        ElasticNetParameters::default()
            .with_tol(1e-10)
            .with_normalize(true)
            .with_alpha(1.0)
            .with_max_iter(10_000)
            .with_l1_ratio(0.5),
    )
    .with_knn_regressor_settings(
        KNNRegressorParameters::default()
            .with_algorithm(KNNAlgorithmName::CoverTree)
            .with_k(3)
            .with_distance(Distance::Euclidean)
            .with_weight(KNNWeightFunction::Uniform),
    )
    .with_svr_settings(
        SVRParameters::default()
            .with_eps(1e-10)
            .with_tol(1e-10)
            .with_c(1.0)
            .with_kernel(Kernel::Linear),
    )
    .with_random_forest_regressor_settings(
        RandomForestRegressorParameters::default()
            .with_m(100)
            .with_max_depth(5)
            .with_min_samples_leaf(20)
            .with_n_trees(100)
            .with_min_samples_split(20),
    )
    .with_decision_tree_regressor_settings(
        DecisionTreeRegressorParameters::default()
            .with_min_samples_split(20)
            .with_max_depth(5)
            .with_min_samples_leaf(20),
    );

§Complete classification customization

use automl::settings::{
    Algorithm, CategoricalNBParameters, DecisionTreeClassifierParameters, Distance,
    GaussianNBParameters, KNNAlgorithmName, KNNClassifierParameters, KNNWeightFunction, Kernel,
    LogisticRegressionParameters, LogisticRegressionSolverName, Metric,
    RandomForestClassifierParameters, SVCParameters,
};

let settings = automl::Settings::default_classification()
    .with_number_of_folds(3)
    .shuffle_data(true)
    .verbose(true)
    .skip(Algorithm::RandomForestClassifier)
    .sorted_by(Metric::Accuracy)
    .with_random_forest_classifier_settings(
        RandomForestClassifierParameters::default()
            .with_m(100)
            .with_max_depth(5)
            .with_min_samples_leaf(20)
            .with_n_trees(100)
            .with_min_samples_split(20),
    )
    .with_logistic_settings(
        LogisticRegressionParameters::default()
            .with_alpha(1.0)
            .with_solver(LogisticRegressionSolverName::LBFGS),
    )
    .with_svc_settings(
        SVCParameters::default()
            .with_epoch(10)
            .with_tol(1e-10)
            .with_c(1.0)
            .with_kernel(Kernel::Linear),
    )
    .with_decision_tree_classifier_settings(
        DecisionTreeClassifierParameters::default()
            .with_min_samples_split(20)
            .with_max_depth(5)
            .with_min_samples_leaf(20),
    )
    .with_knn_classifier_settings(
        KNNClassifierParameters::default()
            .with_algorithm(KNNAlgorithmName::CoverTree)
            .with_k(3)
            .with_distance(Distance::Euclidean)
            .with_weight(KNNWeightFunction::Uniform),
    )
    .with_gaussian_nb_settings(GaussianNBParameters::default().with_priors(vec![1.0, 1.0]))
    .with_categorical_nb_settings(CategoricalNBParameters::default().with_alpha(1.0));

Re-exports§

pub use settings_struct::Settings;

Structs§

CategoricalNBParameters: Parameters for categorical naive bayes (re-export from Smartcore) CategoricalNB parameters. Use Default::default() for default values.
DecisionTreeClassifierParameters: Parameters for decision tree classification (re-export from Smartcore) Parameters of Decision Tree
DecisionTreeRegressorParameters: Parameters for decision tree regression (re-export from Smartcore) Parameters of Regression Tree
ElasticNetParameters: Parameters for elastic net regression (re-export from Smartcore) Elastic net parameters
GaussianNBParameters: Parameters for Gaussian naive bayes (re-export from Smartcore) GaussianNB parameters. Use Default::default() for default values.
KNNClassifierParameters: Parameters for k-nearest neighbors (KNN) classification
KNNRegressorParameters: Parameters for k-nearest neighbor (KNN) regression
LassoParameters: Parameters for LASSO regression (re-export from Smartcore) Lasso regression parameters
LinearRegressionParameters: Parameters for linear regression (re-export from Smartcore) Linear Regression parameters
LogisticRegressionParameters: Parameters for logistic regression (re-export from Smartcore) Logistic Regression parameters
RandomForestClassifierParameters: Parameters for random forest classification (re-export from Smartcore) Parameters of the Random Forest algorithm. Some parameters here are passed directly into base estimator.
RandomForestRegressorParameters: Parameters for random forest regression (re-export from Smartcore) Parameters of the Random Forest Regressor Some parameters here are passed directly into base estimator.
RidgeRegressionParameters: Parameters for ridge regression (re-export from Smartcore) Ridge Regression parameters
SVCParameters: Parameters for support vector classification
SVRParameters: Parameters for support vector regression

Enums§

Algorithm: Algorithm options
Distance: Distance metrics
FinalModel: Final model approach
KNNAlgorithmName: Search algorithms for k-nearest neighbor (KNN) regression (re-export from Smartcore) Both, KNN classifier and regressor benefits from underlying search algorithms that helps to speed up queries. KNNAlgorithmName maintains a list of supported search algorithms, see KNN algorithms
KNNWeightFunction: Weighting functions for k-nearest neighbor (KNN) regression (re-export from Smartcore) Weight function that is used to determine estimated value.
Kernel: Kernel options for use with support vector machines
LinearRegressionSolverName: Solvers for linear regression (re-export from Smartcore) Approach to use for estimation of regression coefficients. QR is more efficient but SVD is more stable.
LogisticRegressionSolverName: Parameters for logistic regression (re-export from Smartcore) Solver options for Logistic regression. Right now only LBFGS solver is supported.
Metric: Metrics for evaluating algorithms
PreProcessing: Options for pre-processing the data
RidgeRegressionSolverName: Solvers for ridge regression (re-export from Smartcore) Approach to use for estimation of regression coefficients. Cholesky is more efficient but SVD is more stable.

Module settingsCopy item path

§Settings customization

§Complete regression customization

§Complete classification customization

Re-exports§

Structs§

Enums§

Module settings