给定一个一维下标数组:
a = array([1, 0, 3])
我想把它编码成一个2D数组:
b = array([[0,1,0,0], [1,0,0,0], [0,0,0,1]])
给定一个一维下标数组:
a = array([1, 0, 3])
我想把它编码成一个2D数组:
b = array([[0,1,0,0], [1,0,0,0], [0,0,0,1]])
当前回答
对于1-hot-encoding
one_hot_encode=pandas.get_dummies(array)
例如
享受编码
其他回答
P是一个2d ndarray。 我们想知道哪一个值在一行中是最大的,在这里是1,其他地方是0。
干净简单的解决方案:
max_elements_i = np.expand_dims(np.argmax(p, axis=1), axis=1)
one_hot = np.zeros(p.shape)
np.put_along_axis(one_hot, max_elements_i, 1, axis=1)
我添加了一个简单的补全函数,只使用numpy操作符:
def probs_to_onehot(output_probabilities):
argmax_indices_array = np.argmax(output_probabilities, axis=1)
onehot_output_array = np.eye(np.unique(argmax_indices_array).shape[0])[argmax_indices_array.reshape(-1)]
return onehot_output_array
它以一个概率矩阵作为输入:例如:
[[0.03038822 0.65810204 0.16549407 0.3797123] . [0.02771272 0.2760752 0.3280924 0.33458805]
它会返回
[[0 0 0 0]... [0 0 0 1]
对于1-hot-encoding
one_hot_encode=pandas.get_dummies(array)
例如
享受编码
def one_hot(n, class_num, col_wise=True):
a = np.eye(class_num)[n.reshape(-1)]
return a.T if col_wise else a
# Column for different hot
print(one_hot(np.array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 9, 9, 9, 9, 8, 7]), 10))
# Row for different hot
print(one_hot(np.array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 9, 9, 9, 9, 8, 7]), 10, col_wise=False))
下面是我根据上面的答案和我自己的用例写的一个示例函数:
def label_vector_to_one_hot_vector(vector, one_hot_size=10):
"""
Use to convert a column vector to a 'one-hot' matrix
Example:
vector: [[2], [0], [1]]
one_hot_size: 3
returns:
[[ 0., 0., 1.],
[ 1., 0., 0.],
[ 0., 1., 0.]]
Parameters:
vector (np.array): of size (n, 1) to be converted
one_hot_size (int) optional: size of 'one-hot' row vector
Returns:
np.array size (vector.size, one_hot_size): converted to a 'one-hot' matrix
"""
squeezed_vector = np.squeeze(vector, axis=-1)
one_hot = np.zeros((squeezed_vector.size, one_hot_size))
one_hot[np.arange(squeezed_vector.size), squeezed_vector] = 1
return one_hot
label_vector_to_one_hot_vector(vector=[[2], [0], [1]], one_hot_size=3)