Numpy：改善数组的奇特索引

Question

Looking for faster fancy indexing for numpy, the code I am running slows down, at np.take() . I tried order=F/C with np.reshape() , no improvement. Python operator works well without the double transpose , but with them is equal to np.take().

p    = np.random.randn(3500, 51)
rows = np.asarray(range(p.shape[0]))
cols = np.asarray([1,2,3,4,5,6,7,8,9,10,15,20,25,30,40,50])

%timeit p[rows][:, cols]
%timeit p.take(cols, axis = 1 )
%timeit np.asarray(operator.itemgetter(*cols)(p.T)).T

1000 loops, best of 3: 301 µs per loop
10000 loops, best of 3: 132 µs per loop
10000 loops, best of 3: 135 µs per loop

Answer 1

A test of several options:

In [3]: p[rows][:,cols].shape
Out[3]: (3500, 16)
In [4]: p[rows[:,None],cols].shape
Out[4]: (3500, 16)
In [5]: p[:,cols].shape
Out[5]: (3500, 16)
In [6]: p.take(cols,axis=1).shape
Out[6]: (3500, 16)

time tests - plain p[:,cols] is fastest. Use a slice where possible.

In [7]: timeit p[rows][:,cols].shape
100 loops, best of 3: 2.78 ms per loop
In [8]: timeit p.take(cols,axis=1).shape
1000 loops, best of 3: 739 µs per loop
In [9]: timeit p[rows[:,None],cols].shape
1000 loops, best of 3: 1.43 ms per loop
In [10]: timeit p[:,cols].shape
1000 loops, best of 3: 649 µs per loop

I've seen itemgetter used for lists, but not arrays. It's a class that iterates of a set of indexes. These 2 lines are doing the same thing:

In [23]: timeit np.asarray(operator.itemgetter(*cols)(p.T)).T.shape
1000 loops, best of 3: 738 µs per loop
In [24]: timeit np.array([p.T[c] for c in cols]).T.shape
1000 loops, best of 3: 748 µs per loop

Notice that pT[c] is pT[c,:] or p[:,c].T . With relatively few cols , and by ignoring advanced indexing with rows , it times close to p[:,cols] .