Examples

One of the best ways to learn how to do new things is to look at examples of how other people did them. To that effect, we’ve included little examples of how you can do stuff with NumPy, both in a straightforward way and in an optimized way. To quantify the effects of these optimization, we’ll use the Timer object we discussed in the Getting Started guide.

Filling a Matrix

We’re going to fill a 10x10 matrix with numbers from 0...99.

Straightforward Way

We can just loop through and provide a number for each position

import numpy as np
mat = np.zeros((10, 10))
for i in range(10):
    for j in range(10):
        mat[i, j] = i*10 + j

Optimized Way

This can be done much more easily using np.arange and np.reshape

import numpy as np
mat = np.reshape(np.arange(100), (10, 10))

Timing Comparison

At small sizes, both are fast

import numpy as np
with Timer("Straightforward Way"):
    mat = np.zeros((10, 10))
    for i in range(10):
        for j in range(10):
            mat[i, j] = i * 10 + j

with Timer("Optimized Way"):
    mat = np.reshape(np.arange(100), (10, 10))
## Out:
# Straightforward Way: took 7e-05s
# Optimized Way: took 4e-05s

as we scale up the performance difference becomes significant

import numpy as np
with Timer("Straightforward Way"):
    mat = np.zeros((1000, 1000))
    for i in range(1000):
        for j in range(1000):
            mat[i, j] = i * 1000 + j

with Timer("Optimized Way"):
    mat = np.reshape(np.arange(1000*1000), (1000, 1000))
## Out:
# Straightforward Way: took 0.42455s
# Optimized Way: took 0.00453s

we’re two orders of magnitude faster if we avoid the loop

Contributing

For now, we’re just collecting a list of examples. Once we have enough we’ll impose some structure on them. Feel free to edit this and add at your leisure, and later we’ll add some jump links to the top of the page once the scrolling gets to be too much.

Edit on GitHub