[Numpy-discussion] finding close together points.
Tue Nov 10 19:22:11 CST 2009
On Tue, Nov 10, 2009 at 8:17 PM, Christopher Barker
> James Bergstra wrote:
>> In some cases a brute-force approach is also good.
>> If r is a matrix of shape Nx2:
>> (r*r).sum(axis=1) -2 * numpy.dot(r, r.T) +
>> (r*r).sum(axis=1).reshape((r.shape, 1)) < thresh**2
>> It's brute force, but it takes advantage of fast matrix multiplication.
> I'm more concerned about memory -- doesn't this use N^2 memory? Which
> could be an issue here.
Yes, this uses N^2 time and space. It's not a good algorithm when N is large.
More information about the NumPy-Discussion