Vectorize Loop Using Numpy

Question

I have two 1D Numpy arrays index_data and charge, which are the same length and contain ints and floats respectively. I am trying to make a total_charge array as follows:

total_charge = []
for i in range(len(index_data)):
    if index_data[i] == 0:
        total_charge.append(charge[i])
    else:
        total_charge[-1] += charge[i]
total_charge = np.array(total_charge)

How would I vectorize this? Help me Numpy wizards, you're my only hope.

I don't think this lends itself to vectorization very easily ... — mgilson
– mgilson, Commented May 22, 2014 at 19:52
A Star Wars reference from Mr. Spock? No, that's just not right. :) — Warren Weckesser
– Warren Weckesser, Commented May 22, 2014 at 20:27

Warren Weckesser · Accepted Answer · 2014-05-22 20:20:49Z

4

A concrete example of index_data and charge would help clarify the expected input. From reading what you have, however, I think using np.add.reduceat can be used.

Apparently each new group of charges is indicated by a 0 in index_data; otherwise, the values in index_data are not used. total_charge is the sum of the charges in each group.

Here's an example of how reduceat could be used.

First, some demo data:

In [105]: index_data
Out[105]: array([0, 1, 1, 0, 1, 1, 1, 0, 1])

In [106]: charge
Out[106]: array([ 1.5,  2. ,  3. ,  2.5,  1.5,  1. ,  1. ,  2. ,  1. ])

zerolocs gives the indices where 0 occurs in index_data:

In [107]: zerolocs = where(index_data==0)[0]

In [108]: zerolocs
Out[108]: array([0, 3, 7])

Use np.add.reduceat to sum the charge groups.

In [109]: total_charge = np.add.reduceat(charge, zerolocs)

In [110]: total_charge
Out[110]: array([ 6.5,  6. ,  3. ])

answered May 22, 2014 at 20:20

Warren Weckesser

116k20 gold badges207 silver badges224 bronze badges

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

1 Comment

Jacob Garber Over a year ago

You are a miracle worker. I knew there was some obscure Numpy function that would do exactly what I wanted, I just didn't know which one. And hey, us Trekkies can still appreciate Star Wars.

Collectives™ on Stack Overflow

Vectorize Loop Using Numpy

1 Answer 1

1 Comment

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

1 Comment

Your Answer

Sign up or log in

Post as a guest

Related