mlektic.logistic_reg package

Submodules

mlektic.logistic_reg.logistic_regression_archt module

class mlektic.logistic_reg.logistic_regression_archt.LogisticRegressionArcht(iterations: int = 50, use_intercept: bool = True, verbose: bool = True, regularizer: Callable = None, optimizer: Optional[Tuple[Optimizer, str, int]] = None, method: str = 'mle', metric: str = 'accuracy')

Bases: object

Logistic Regression model class supporting different training methods including Maximum Likelihood Estimation (MLE), batch gradient descent, stochastic gradient descent, and mini-batch gradient descent.

iterations

Number of training iterations (only applicable for gradient descent methods).

Type:

int

use_intercept

Whether to include an intercept in the model.

Type:

bool

verbose

Whether to print training progress.

Type:

bool

weights

Model weights.

Type:

tf.Variable

cost_history

History of cost values during training (only applicable for gradient descent methods).

Type:

list

metric_history

History of metric values during training (only applicable for gradient descent methods).

Type:

list

n_features

Number of features in the input data.

Type:

int

regularizer

Regularization function.

Type:

callable

optimizer

Optimizer for gradient descent.

Type:

tf.optimizers.Optimizer

method

Training method to use.

Type:

str

metric

Evaluation metric to use.

Type:

str

num_classes

Number of classes in the target variable.

Type:

int

train(train_set)

Trains the model on the provided training set.

get_parameters()

Returns the model parameters (weights).

get_intercept()

Returns the model intercept.

get_cost_history()

Returns the history of cost values.

get_metric_history()

Returns the history of metric values.

predict(input_new)

Predicts output for new input data.

eval(test_set, metric)

Evaluates the model on a test set using the specified metric.

save_model(filepath)

Saves the model to a file.

load_model(filepath)

Loads the model from a file.

eval(test_set: Tuple[ndarray, ndarray], metric: str, to_df: bool = True)

Evaluates the model on a test set using the specified metric.

Parameters:

• test_set (tuple) – Tuple containing test input data (np.ndarray) and output data (np.ndarray).

• metric (str) – Metric to use for evaluation. Options are ‘categorical_crossentropy’, ‘binary_crossentropy’, ‘accuracy’, ‘precision’, ‘recall’, ‘f1_score’, ‘confusion_matrix’.

• to_df (bool, optional) – If True and metric is ‘confusion_matrix’, returns the matrix as a pandas DataFrame. If False, returns it as a numpy array. Default is True.

Returns:

The evaluation result. For metrics other than ‘confusion_matrix’,

returns a float. For ‘confusion_matrix’, returns either a DataFrame or a tensor converted to numpy, based on to_df.

Return type:

Union[float, pd.DataFrame, np.ndarray]

Raises:

ValueError – If the specified metric is not supported.

get_cost_history() → list

Returns the history of cost values during training.

Returns:

List of cost values.

Return type:

list

get_intercept() → Optional[ndarray]

Returns the model intercept.

Returns:

Intercept value if use_intercept is True, else None.

Return type:

Union[np.ndarray, None]

get_metric_history() → list

Returns the history of metric values during training.

Returns:

List of metric values.

Return type:

list

get_parameters() → ndarray

Returns the model parameters (weights).

Returns:

Array of model parameters.

Return type:

np.ndarray

load_model(filepath: str) → None

Loads the model from a file.

Parameters:

filepath (str) – Path to the file from which the model will be loaded.

predict_class(input_new: Union[ndarray, list, float]) → ndarray

Predicts the class for new input data.

Parameters:

input_new (Union[np.ndarray, list, float]) – New input data for prediction.

Returns:

Predicted class(es).

Return type:

np.ndarray

Raises:

ValueError – If the input does not have the expected number of features.

predict_prob(input_new: Union[ndarray, list, float]) → ndarray

Predicts output for new input data.

Parameters:

input_new (Union[np.ndarray, list, float]) – New input data for prediction.

Returns:

Predicted probabilities.

Return type:

np.ndarray

Raises:

ValueError – If the input does not have the expected number of features.

save_model(filepath: str) → None

Saves the model to a file.

Parameters:

filepath (str) – Path to the file where the model will be saved.

train(train_set: Tuple[ndarray, ndarray]) → LogisticRegressionArcht

Trains the model on the provided training set.

Parameters:

train_set (tuple) – Tuple containing training input data (np.ndarray) and output data (np.ndarray).

Returns:

The trained model instance.

Return type:

LogisticRegressionArcht

mlektic.logistic_reg.logreg_utils module

mlektic.logistic_reg.logreg_utils.calculate_accuracy(y_true: Tensor, y_pred: Tensor) → Tensor

Calculates the accuracy between true labels and predicted probabilities.

Parameters:

• y_true (tf.Tensor) – True labels, one-hot encoded. Shape should be (n_samples, num_classes).

• y_pred (tf.Tensor) – Predicted probabilities. Shape should be (n_samples, num_classes).

Returns:

Accuracy.

Return type:

tf.Tensor

mlektic.logistic_reg.logreg_utils.calculate_binary_crossentropy(y_true: Tensor, y_pred: Tensor) → Tensor

Calculates the binary cross-entropy loss between true labels and predicted probabilities.

Parameters:

• y_true (tf.Tensor) – True labels. Shape should be (n_samples,).

• y_pred (tf.Tensor) – Predicted probabilities. Shape should be (n_samples, 2).

Returns:

Binary cross-entropy loss.

Return type:

tf.Tensor

mlektic.logistic_reg.logreg_utils.calculate_categorical_crossentropy(y_true: Tensor, y_pred: Tensor) → Tensor

Calculates the categorical cross-entropy loss between true labels and predicted probabilities.

Parameters:

• y_true (tf.Tensor) – True labels, one-hot encoded. Shape should be (n_samples, num_classes).

• y_pred (tf.Tensor) – Predicted probabilities. Shape should be (n_samples, num_classes).

Returns:

Categorical cross-entropy loss.

Return type:

tf.Tensor

mlektic.logistic_reg.logreg_utils.calculate_confusion_matrix(y_true: Tensor, y_pred: Tensor, to_df: bool = True)

Calculates the confusion matrix between true labels and predicted probabilities.

Parameters:

• y_true (tf.Tensor) – True labels, one-hot encoded. Shape should be (n_samples, num_classes).

• y_pred (tf.Tensor) – Predicted probabilities. Shape should be (n_samples, num_classes).

• to_df (bool, optional) – If True, returns the confusion matrix as a formatted pandas DataFrame. If False, returns the tensor. Default is True.

Returns:

Confusion matrix in tensor form (if to_df=False) or

a formatted DataFrame (if to_df=True).

Return type:

Union[tf.Tensor, pd.DataFrame]

mlektic.logistic_reg.logreg_utils.calculate_f1_score(y_true: Tensor, y_pred: Tensor) → Tensor

Calculates the F1 score between true labels and predicted probabilities.

Parameters:

• y_true (tf.Tensor) – True labels, one-hot encoded. Shape should be (n_samples, num_classes).

• y_pred (tf.Tensor) – Predicted probabilities. Shape should be (n_samples, num_classes).

Returns:

F1 score.

Return type:

tf.Tensor

mlektic.logistic_reg.logreg_utils.calculate_precision(y_true: Tensor, y_pred: Tensor) → Tensor

Calculates the precision between true labels and predicted probabilities.

Parameters:

• y_true (tf.Tensor) – True labels, one-hot encoded. Shape should be (n_samples, num_classes).

• y_pred (tf.Tensor) – Predicted probabilities. Shape should be (n_samples, num_classes).

Returns:

Precision.

Return type:

tf.Tensor

mlektic.logistic_reg.logreg_utils.calculate_recall(y_true: Tensor, y_pred: Tensor) → Tensor

Calculates the recall between true labels and predicted probabilities.

Parameters:

• y_true (tf.Tensor) – True labels, one-hot encoded. Shape should be (n_samples, num_classes).

• y_pred (tf.Tensor) – Predicted probabilities. Shape should be (n_samples, num_classes).

Returns:

Recall.

Return type:

tf.Tensor

Module contents