三维混淆矩阵keras代码

下面是使用Keras实现的三维混淆矩阵的代码示例：

import numpy as np
from keras import backend as K

def confusion_matrix_3d(y_true, y_pred):
    """
    Computes a 3D confusion matrix.
    Args:
        y_true: true labels, tensor with shape (batch_size, num_classes, num_samples)
        y_pred: predicted labels, tensor with shape (batch_size, num_classes, num_samples)
    Returns:
        3D confusion matrix, tensor with shape (num_classes, num_classes, num_samples)
    """
    y_true = K.argmax(y_true, axis=1)
    y_pred = K.argmax(y_pred, axis=1)
    num_samples = K.shape(y_true)[-1]
    num_classes = K.int_shape(y_true)[0]
    cm = np.zeros((num_classes, num_classes, num_samples))
    for i in range(num_samples):
        cm[:,:,i] = np.array(K.eval(K.tf.math.confusion_matrix(y_true[:,i], y_pred[:,i])))
    return cm

在使用此函数时，需要将真实值和预测值转换为one-hot编码的形式，以便使用Keras的`argmax`函数进行轴变换。此函数会返回一个形状为`(num_classes, num_classes, num_samples)`的三维混淆矩阵，其中`num_classes`是类别数，`num_samples`是样本数。