【python】矩阵嵌套的简单解释

经常看到numpy或者pytorch出现矩阵的嵌套A[B]，不太懂啥意思，下面转载的这个知乎一句话解释清楚了。
原文链接：https://zhuanlan.zhihu.com/p/570847895

mport numpy as np

png = np.array([[0, 0, 1, 1], [1, 2, 1, 3], [2, 2, 3, 4], [3, 3, 2, 2]])
print(png)
'''
[[0 0 1 1]
 [1 2 1 3]
 [2 2 3 4]
 [3 3 2 2]]
'''
num_classes = 2
png[png >= num_classes] = num_classes
print(png)
'''
[[0 0 1 1]
 [1 2 1 2]
 [2 2 2 2]
 [2 2 2 2]]
'''
print(png.reshape([-1]))
'''
[0 0 1 1 1 2 1 2 2 2 2 2 2 2 2 2]
'''
seg_labels = np.eye(num_classes + 1)[png.reshape([-1])]
print(seg_labels)
'''
[[1. 0. 0.]
 [1. 0. 0.]
 [0. 1. 0.]
 [0. 1. 0.]
 [0. 1. 0.]
 [0. 0. 1.]
 [0. 1. 0.]
 [0. 0. 1.]
 [0. 0. 1.]
 [0. 0. 1.]
 [0. 0. 1.]
 [0. 0. 1.]
 [0. 0. 1.]
 [0. 0. 1.]
 [0. 0. 1.]
 [0. 0. 1.]]
'''

最经典的一句话：矩阵A，和矩阵B，AB表示根据B矩阵的元素去索引矩阵A，得到与矩阵B大小相同的矩阵。np.eye得到一个对角矩阵。这段代码将像素值转为独热编码。