multilogistic.py 文件源码

def _multinomial_loss(w, X, Y, alpha, sample_weight):
    """Computes multinomial loss and class probabilities.

    Parameters
    ----------
    w : ndarray, shape (n_classes * n_features,) or
        (n_classes * (n_features + 1),)
        Coefficient vector.

    X : {array-like, sparse matrix}, shape (n_samples, n_features)
        Training data.

    Y : ndarray, shape (n_samples, n_classes)
        Transformed labels according to the output of LabelBinarizer.

    alpha : float
        Regularization parameter. alpha is equal to 1 / C.

    sample_weight : array-like, shape (n_samples,) optional
        Array of weights that are assigned to individual samples.
        If not provided, then each sample is given unit weight.

    Returns
    -------
    loss : float
        Multinomial loss.

    p : ndarray, shape (n_samples, n_classes)
        Estimated class probabilities.

    w : ndarray, shape (n_classes, n_features)
        Reshaped param vector excluding intercept terms.

    Reference
    ---------
    Bishop, C. M. (2006). Pattern recognition and machine learning.
    Springer. (Chapter 4.3.4)
    """
    n_classes = Y.shape[1]
    n_features = X.shape[1]
    fit_intercept = w.size == (n_classes * (n_features + 1))
    w = w.reshape(n_classes, -1)
    alpha = alpha.reshape(n_classes, -1)
    sample_weight = sample_weight[:, np.newaxis]
    if fit_intercept:
        intercept = w[:, -1]
        w = w[:, :-1]
    else:
        intercept = 0
    p = safe_sparse_dot(X, w.T)
    p += intercept
    p -= logsumexp(p, axis=1)[:, np.newaxis]
    loss = -(sample_weight * Y * p).sum()
    loss += 0.5 * ((alpha + L2_REG) * w * w).sum()
    p = np.exp(p, p)
    return loss, p, w