losses.py 文件源码

def poisson_loss(observed_ratings, predicted_ratings):
    """
    Poisson loss.

    Parameters
    ----------

    observed_ratings: tensor
        Tensor containing observed ratings.
    predicted_ratings: tensor
        Tensor containing rating predictions.

    Returns
    -------

    loss, float
        The mean value of the loss function.
    """

    assert_no_grad(observed_ratings)

    return (predicted_ratings - observed_ratings * torch.log(predicted_ratings)).mean()