Skip to content

Documentation for FRegressionFilterPValue

Feature Selection for Regression - p-value based.

Fit f_regression and select all columns when p_value of column < p_threshold.

__init__(self, p_threshold=0.05) special

Initialize the object.

Parameters:

Name Type Description Default
p_threshold float

Upper bound for p_values.

0.05
Source code in photonai/modelwrapper/feature_selection.py
def __init__(self, p_threshold: float = .05):
    """
    Initialize the object.

    Parameters:
        p_threshold:
            Upper bound for p_values.

    """
    self.p_threshold = p_threshold
    self.selected_indices = []
    self.n_original_features = None

fit(self, X, y)

Calculation of the important columns.

Apply f_regression on input X, y to generate p_values. selected_indices = all p_value(columns) < p_threshold.

Parameters:

Name Type Description Default
X ndarray

The input samples of shape [n_samples, n_original_features]

required
y ndarray

The input targets of shape [n_samples, 1]

required
Source code in photonai/modelwrapper/feature_selection.py
def fit(self, X: np.ndarray, y: np.ndarray):
    """Calculation of the important columns.

    Apply f_regression on input X, y to generate p_values.
    selected_indices = all p_value(columns) < p_threshold.

    Parameters:
        X:
            The input samples of shape [n_samples, n_original_features]

        y:
            The input targets of shape [n_samples, 1]

    """
    self.n_original_features = X.shape[1]
    _, p_values = f_regression(X, y)
    self.selected_indices = np.where(p_values < self.p_threshold)[0]
    return self

inverse_transform(self, X)

Reverse to original dimension.

Parameters:

Name Type Description Default
X ndarray

The input samples of shape [n_samples, n_selected_features].

required

Exceptions:

Type Description
ValueError

If input X has a different shape than during fitting.

Returns:

Type Description
ndarray

Array of shape [n_samples, n_original_features] with columns of zeros inserted where features would have been removed.

Source code in photonai/modelwrapper/feature_selection.py
def inverse_transform(self, X: np.ndarray) -> np.ndarray:
    """Reverse to original dimension.

    Parameters:
        X:
            The input samples of shape [n_samples, n_selected_features].

    Raises:
        ValueError: If input X has a different shape than during fitting.

    Returns:
        Array of shape [n_samples, n_original_features]
        with columns of zeros inserted where features would have
        been removed.

    """
    if X.shape[1] != len(self.selected_indices):
        msg = "X has a different shape than during fitting."
        logger.error(msg)
        raise ValueError(msg)

    Xt = np.zeros((X.shape[0], self.n_original_features))
    Xt[:, self.selected_indices] = X
    return Xt

transform(self, X)

Reduced input X to selected_columns.

Returns:

Type Description
ndarray

Column-filtered array of shape [n_samples, n_selected_features].

Source code in photonai/modelwrapper/feature_selection.py
def transform(self, X: np.ndarray) -> np.ndarray:
    """Reduced input X to selected_columns.

    Parameters:
        X
            The input samples of shape [n_samples, n_original_features]

    Returns:
        Column-filtered array of shape [n_samples, n_selected_features].

    """
    return X[:, self.selected_indices]