focal损失函数的python代码

Focal Loss是一种用于处理类别不平衡问题的损失函数。以下是Focal Loss的Python代码实现：

import tensorflow as tf

def focal_loss(y_true, y_pred, alpha=0.25, gamma=2.0):
    """
    Focal loss for multi-class classification
    :param y_true: ground truth labels, shape of [batch_size, num_cls]
    :param y_pred: model's output, shape of [batch_size, num_cls]
    :param alpha: balancing factor, default to be 0.25
    :param gamma: focusing parameter, default to be 2.0
    :return: focal loss
    """
    # transform y_true to one-hot encoding
    y_true = tf.one_hot(tf.cast(y_true, tf.int32), depth=tf.shape(y_pred)[1])

    # clip y_pred to prevent NaN when taking log of it
    y_pred = tf.clip_by_value(y_pred, 1e-8, 1.0)

    # calculate focal loss
    pt = tf.where(tf.equal(y_true, 1), y_pred, 1 - y_pred)
    focal_loss = - alpha * tf.pow(1 - pt, gamma) * tf.math.log(pt)

    return tf.reduce_mean(focal_loss)

其中，`y_true`是真实标签，`y_pred`是模型预测输出，`alpha`是平衡因子，`gamma`是聚焦参数。函数首先将`y_true`转换为one-hot编码，然后将`y_pred`限制在（0，1）之间以避免取log时出现NaN。接着计算Focal Loss，并返回其均值。