如何使用Keras计算马氏距离？

import numpy as np
from keras import backend as K

# 准备数据
data = np.array([[1, 2, 3], [4, 5, 6], [7, 8, 9]])

# 数据预处理
data = (data - np.mean(data, axis=0)) / np.std(data, axis=0)

# 计算协方差矩阵
cov_matrix = np.cov(data, rowvar=False)

# 定义马氏距离函数
def mahalanobis_distance(x, y):
    diff = K.expand_dims(x - y, axis=0)
    inv_cov = K.constant(np.linalg.inv(cov_matrix))
    mahalanobis = K.sqrt(K.dot(K.dot(diff, inv_cov), K.transpose(diff)))
    return mahalanobis

# 创建输入张量
input_x = K.placeholder(shape=(3,))
input_y = K.placeholder(shape=(3,))

# 计算马氏距离
distance = mahalanobis_distance(input_x, input_y)

# 创建Keras函数
calculate_distance = K.function([input_x, input_y], [distance])

# 使用示例
sample_x = np.array([1, 2, 3])
sample_y = np.array([4, 5, 6])
result = calculate_distance([sample_x, sample_y])

print("马氏距离:", result)