autoqild.automl.model_configurations

Configurations for search space for AutoGluon tools.

Module Attributes

hyperparameters	This dictionary defines the hyperparameters for several models like NN_TORCH, GBM, CAT, XGB, FASTAI, RF, XT, and KNN.
reduced_hyperparameters	This dictionary defines the hyperparameters for simpler models like FASTAI, RF and XT.

autoqild.automl.model_configurations.hyperparameters = {'CAT': {'depth': Int: lower=4, upper=10, 'l2_leaf_reg': Real: lower=0.1, upper=10, 'learning_rate': Real: lower=0.01, upper=0.5}, 'FASTAI': {'emb_drop': Real: lower=0.0, upper=0.5, 'learning_rate': Real: lower=1e-05, upper=0.1, 'ps': Real: lower=0.0, upper=0.5, 'smoothing': Real: lower=0.0, upper=0.5, 'wd': Real: lower=1e-06, upper=0.1}, 'GBM': {'bagging_fraction': Real: lower=0.2, upper=0.95, 'feature_fraction': Real: lower=0.2, upper=0.95, 'lambda_l1': Real: lower=1e-06, upper=0.01, 'lambda_l2': Real: lower=1e-06, upper=0.01, 'learning_rate': Real: lower=0.01, upper=0.5, 'max_depth': Int: lower=3, upper=20, 'min_data_in_leaf': Int: lower=20, upper=5000, 'n_estimators': Int: lower=20, upper=300, 'num_leaves': Int: lower=20, upper=300}, 'KNN': {'n_neighbors': Int: lower=3, upper=5, 'p': Categorical[1, 2, 3], 'weights': Categorical['uniform', 'distance']}, 'NN_TORCH': {'dropout_prob': Real: lower=0.0, upper=0.5, 'hidden_size': Int: lower=8, upper=256, 'learning_rate': Real: lower=1e-05, upper=0.1, 'num_layers': Int: lower=2, upper=20}, 'RF': {'class_weight': Categorical['balanced', 'balanced_subsample'], 'criterion': Categorical['gini', 'entropy'], 'max_depth': Int: lower=6, upper=20, 'max_features': Categorical['sqrt', 'log2'], 'min_samples_leaf': Int: lower=2, upper=50, 'min_samples_split': Int: lower=2, upper=50, 'n_estimators': Int: lower=20, upper=300}, 'XGB': {'gamma': Real: lower=0.0, upper=1.0, 'learning_rate': Real: lower=0.01, upper=0.5, 'max_depth': Int: lower=3, upper=10, 'n_estimators': Int: lower=20, upper=300, 'reg_alpha': Real: lower=0.0, upper=1.0, 'reg_lambda': Real: lower=0.0, upper=1.0}, 'XT': {'class_weight': Categorical['balanced', 'balanced_subsample'], 'criterion': Categorical['gini', 'entropy'], 'max_depth': Int: lower=6, upper=20, 'max_features': Categorical['sqrt', 'log2'], 'min_samples_leaf': Int: lower=2, upper=50, 'min_samples_split': Int: lower=2, upper=50, 'n_estimators': Int: lower=20, upper=300}}

This dictionary defines the hyperparameters for several models like NN_TORCH, GBM, CAT, XGB, FASTAI, RF, XT, and KNN.

These models are commonly used in machine learning pipelines, and each hyperparameter is configured using the Real, Int, or Categorical space from AutoGluon, which supports hyperparameter tuning.

autoqild.automl.model_configurations.reduced_hyperparameters = {'FASTAI': {'emb_drop': Real: lower=0.0, upper=0.5, 'learning_rate': Real: lower=1e-05, upper=0.1, 'ps': Real: lower=0.0, upper=0.5, 'smoothing': Real: lower=0.0, upper=0.5, 'wd': Real: lower=1e-06, upper=0.1}, 'RF': {'class_weight': Categorical['balanced'], 'criterion': Categorical['gini', 'entropy'], 'max_depth': Int: lower=6, upper=20, 'max_features': Categorical['sqrt', 'log2'], 'min_samples_leaf': Int: lower=2, upper=50, 'min_samples_split': Int: lower=2, upper=50, 'n_estimators': Int: lower=20, upper=300}, 'XT': {'class_weight': Categorical['balanced'], 'criterion': Categorical['gini', 'entropy'], 'max_depth': Int: lower=6, upper=20, 'max_features': Categorical['sqrt', 'log2'], 'min_samples_leaf': Int: lower=2, upper=50, 'min_samples_split': Int: lower=2, upper=50, 'n_estimators': Int: lower=20, upper=300}}

This dictionary defines the hyperparameters for simpler models like FASTAI, RF and XT.

These models are commonly used in machine learning pipelines, and each hyperparameter is configured using the Real, Int, or Categorical space from AutoGluon, which supports hyperparameter tuning.