我有一组范围从3到27的值,它们有20个有限值:
A = [(0,21),(1,12),(2,15),(3,3),(4,21),(5,15),(6,27),(7,21),(8,9),(9,27),(10,12),(11,9),(12,12),(13,3),(14,9),(15,12),(16,6),(17,3),(18,9),(19,15)]
我想学习如何创建一个具有9个bin的numpy数组,每个bin对于给定的元组[1]具有-1和+1整数的范围,3的范围为3的倍数27(但这应该可以与整数和范围的任何其他组合互换)。最后,我希望能够创建一个看起来像这样的矩阵:
[[0,0,0,0,0,0,1,0,0],
[0,0,0,1,0,0,0,0,0],
[0,0,0,0,1,0,0,0,0],
[1,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,1,0,0],
[0,0,0,0,1,0,0,0,0],
[0,0,0,0,0,0,0,0,1],
....]
我正在阅读numpy如何(num, bins) = histogram(x, bins=None, range=None)
,但我不太清楚该怎么做。
我原以为我必须遍历'A'来获取唯一值('a'),然后通过(a-1,a + 1)进行范围,以得到我想要的箱数只到len(unique_values)
。但后来我迷路了。谁能指导我?
答案 0 :(得分:2)
这是np.searchsorted
/ np.digitize
-
bins = np.arange(3,28,3)
ar = np.asarray(A)[:,1] # or np.array([i[1] for i in A])
ids = np.searchsorted(bins, ar) # or np.digitize(ar,bins)-1
out = (ids[:,None] == np.arange(9)).astype(int)
获取最终输出的最后一步可以替换为array-initialization -
out = np.zeros((len(ids), 9),dtype=int)
out[np.arange(len(ids)), ids] = 1
如果元组中的第一个元素不是按顺序排列的,我们可能希望使用它们来索引到行 -
out[np.asarray(A)[:,0], ids] = 1
示例运行 -
In [205]: A
Out[205]:
[(0, 21),
(1, 12),
(2, 15),
(3, 3),
(4, 21),
(5, 15),
(6, 27),
(7, 21),
(8, 9),
(9, 27),
(10, 12),
(11, 9),
(12, 12),
(13, 3),
(14, 9),
(15, 12),
(16, 6),
(17, 3),
(18, 9),
(19, 15)]
In [206]: out[:7] # first 7 rows of output
Out[206]:
array([[0, 0, 0, 0, 0, 0, 1, 0, 0],
[0, 0, 0, 1, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 1, 0, 0, 0, 0],
[1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 1, 0, 0],
[0, 0, 0, 0, 1, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 1]])