Numpy index a 1D array to 2D

Question

Below is code with the functionality I want on some simple sample data. Basically I binned data using np.digitize, and then I computed a column index based on this question. bin_idx is known to never decrease in case that helps. How can I index to get the 2D array without an explicit loop? One complication is that the number of values in each row/bin varies. I will later do different statistics on each bin/row, max just being an example.

import numpy as np

x = np.arange(10)
bin_idx = np.array([0, 0, 0, 1, 2, 3, 3, 4, 4, 4])
col_idx = np.array([0, 1, 2, 0, 0, 0, 1, 0, 1, 2])

binned = np.ones((bin_idx[-1]+1, np.max(col_idx)+1)) * np.nan
for i in range(len(x)):
    binned[bin_idx[i], col_idx[i]] = x[i]
print(binned)
row_max = np.nanmax(binned, 1)
print(row_max)

Aside: if you're working with data, pandas might be more natural than working with bare numpy; here, you're basically reimplementing something like df.pivot(1,2,0).max(axis=1). — DSM
– DSM, Commented Feb 25, 2016 at 16:16

Alex · Accepted Answer · 2016-02-25 16:14:09Z

3

Numpy indexing allows you to pass sequences as indices. Also check out Numpy's full method used below to create the binned array.

binned = np.full((bin_idx[-1]+1, np.max(col_idx)+1), np.nan)
binned[bin_idx, col_idx] = x

edited Feb 25, 2016 at 16:14

answered Feb 25, 2016 at 16:08

Alex

19.2k9 gold badges65 silver badges82 bronze badges

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

Collectives™ on Stack Overflow

Numpy index a 1D array to 2D

1 Answer 1

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related