multilogistic.py 文件源码

def _multinomial_loss_grad(w, X, Y, alpha, sample_weight):
    """Computes the multinomial loss, gradient 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.

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

    grad : ndarray, shape (n_classes * n_features,) or
        (n_classes * (n_features + 1),)
        Ravelled gradient of the multinomial loss.

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

    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))
    grad = np.zeros((n_classes, n_features + bool(fit_intercept)))
    alpha = alpha.reshape(n_classes, -1)
    loss, p, w = _multinomial_loss(w, X, Y, alpha, sample_weight)
    sample_weight = sample_weight[:, np.newaxis]
    diff = sample_weight * (p - Y)
    grad[:, :n_features] = safe_sparse_dot(diff.T, X)
    grad[:, :n_features] += (alpha + L2_REG) * w
    if fit_intercept:
        grad[:, -1] = diff.sum(axis=0)
    return loss, grad.ravel(), p