向量化参数不同维的函数

Question

我正在尝试根据NumPy中的均值和协方差矩阵创建一系列高斯。

import numpy as np
from scipy.stats import multivariate_normal

means = np.array([
        [ 1,  0],
        [ 2,  2],
        [ 3, -1]])
covs = np.array([
       [[1, 0],
        [0, 1]],
       [[9, 0],
        [0, 4]],
       [[4, 1],
        [1, 1]]])

# This works
gaussians = [multivariate_normal(mean, cov) for mean, cov in zip(means, covs)]
'''
[<scipy.stats._multivariate.multivariate_normal_frozen at 0x27ced7e9f60>,
 <scipy.stats._multivariate.multivariate_normal_frozen at 0x27ced7e92e8>,
 <scipy.stats._multivariate.multivariate_normal_frozen at 0x27ced7e92b0>]
'''

# This throws an error 
vec_gaussian = np.vectorize(lambda mean, cov: multivariate_normal(mean, cov))
gaussians = vec_gaussian(means, covs)
''' ValueError: operands could not be broadcast together with shapes (3,2) (3,2,2) '''

# Adding an axis gives singular matrix error (???)
gaussians = vec_gaussian(means[:, :, np.newaxis], covs)
''' LinAlgError: singular matrix '''

有什么方法可以使用vectorize还是我最好只使用for循环？