给定ndarray x和包含x维度的连续切片长度的一维数组,我想计算一个包含所有切片之和的新数组.例如,在二维上总结一维:
>>> lens = np.array([1, 3, 2])
array([1, 3, 2])
>>> x = np.arange(4 * lens.sum()).reshape((4, lens.sum())).astype(float)
array([[ 0., 1., 2., 3., 4., 5.],
[ 6., 7., 8., 9., 10., 11.],
[ 12., 13., 14., 15., 16., 17.],
[ 18., 19., 20., 21., 22., 23.]])
# I want to compute:
>>> result
array([[ 0., 6., 9.],
[ 6., 24., 21.],
[ 12., 42., 33.],
[ 18., 60., 45.]])
# 0 = 0
# 6 = 1 + 2 + 3
# ...
# 45 = 22 + 23
想到的两种方式是:
a)使用cumsum和fancy索引:
def cumsum_method(x, lens):
xc = x.cumsum(1)
lc = lens.cumsum() - 1
res = xc[:, lc]
res[:, 1:] -= xc[:, lc[:-1]]
return res
b)使用bincount并智能地生成适当的bin:
def bincount_method(x, lens):
bins = np.arange(lens.size).repeat(lens) + \
np.arange(x.shape[0])[:, None] * lens.size
return np.bincount(bins.flat, weights=x.flat).reshape((-1, lens.size))
在大输入上对这两个进行定时使得cumsum方法表现稍好一些:
>>> lens = np.random.randint(1, 100, 100)
>>> x = np.random.random((100000, lens.sum()))
>>> %timeit cumsum_method(x, lens)
1 loops, best of 3: 3 s per loop
>>> %timeit bincount_method(x, lens)
1 loops, best of 3: 3.9 s per loop
我错过了一种明显更有效的方式吗?看起来本机c调用会更快,因为它不需要分配cumsum或bin数组.一个接近这个的numpy内置函数可能比(a)或(b)更好.我通过搜索和浏览文档找不到任何东西.
注意,这类似于this question,但总和间隔不规则.