Scipy: Partition array into 3 subarrays

Question

I am trying to figure out whether there's a numpy/scipy function to efficiently partition an array into subarrays using a certain rule.

My problem is the following: I have a nxn matrix, lets call it W. And I have a vector h. I now want to partition the column vectors of W into 3 arrays:

W_pos, where > 0 for all vectors w from W_pos
W_null, where = 0 for all vectors w from W_null
W_neg, where < 0 for all vectors w from W_neg

Right now I am doing it like this, which is working but I think it is not very efficient:

    nonzero_indices = (sp.isclose(sp.dot(h_k.T, W),0, 10e-12) == False)
    self.W_null = W[:,~nonzero_indices]

    W_nonzero = W[:,nonzero_indices]
    pos_indices = (sp.dot(h_k.T, W_nonzero) > 0)
    W_pos = W_nonzero[:,pos_indices]
    W_neg = W_nonzero[:,~pos_indices]

Is there a better way? Thanks for your help and if there something not clear please let me know. Cheers

Where's the inefficiency? Generating the masks, or in indexing? — hpaulj
– hpaulj, Commented Nov 30, 2015 at 17:05

Lee · Accepted Answer · 2015-11-30 16:19:32Z

2

w=np.random.random((10,10))-0.5 # example array

.

wneg = w[w<0]

wzero = w[w==0]

wpos = w[w>0]

edited Nov 30, 2015 at 16:19

answered Nov 30, 2015 at 15:28

Lee

31.4k31 gold badges124 silver badges187 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

Doc Over a year ago

Thanks for your answer. How is this faster than my version? In this case I would have to go through the entire array 3 times, once for each partition(pos/neg/zero).

hpaulj Over a year ago

You have to do timings. At this level of coding, reasoning only gets you so far. I suspect indexing takes longer than generating the masks.

Collectives™ on Stack Overflow

Scipy: Partition array into 3 subarrays

1 Answer 1

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related