Skip to content

Added a cubic spline interpolator #6

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Closed
pelson wants to merge 1 commit into from
Closed

Added a cubic spline interpolator #6

pelson wants to merge 1 commit into from

Conversation

@pelson
Copy link
Member

@pelson pelson commented Jan 31, 2017

Surprisingly, this approach turns out to be ~4x faster than

cs = scipy.interpolate.CubicSpline(x, y)
ys = cs(xs)

Though the comparison is really apples vs oranges, as the algorithm implemented is different.

There is loads of room for improvement (and potentially even a better algorithm - this is the best I could find in a short search), but this is good enough to be useful already.

@pelson
Copy link
Member Author

pelson commented Jan 31, 2017

Hmmm. Seems to be suffering from over-fitting. May be the algorithm itself... 😞

array([  9.28148594e+04,   9.24955625e+04,   9.20495625e+04,
         9.14780625e+04,   9.07826562e+04,   8.99653906e+04,
         8.90285156e+04,   8.79746719e+04,   8.68069688e+04,
         8.55286875e+04,   8.41435469e+04,   8.26555156e+04,
         8.10689062e+04,   7.93884062e+04,   7.76193594e+04,
         7.57674844e+04,   7.38387656e+04,   7.18393594e+04,
         6.97743281e+04,   6.76511406e+04,   6.54768594e+04,
         6.32554688e+04,   6.09930781e+04,   5.86990938e+04,
         5.63856250e+04,   5.40647812e+04,   5.17401719e+04,
         4.94096406e+04,   4.70854688e+04,   4.47756719e+04,
         4.24794688e+04,   4.02053906e+04,   3.79599375e+04,
         3.57455156e+04,   3.35653594e+04,   3.14246094e+04,
         2.93292969e+04,   2.72865312e+04,   2.53046094e+04,
         2.33863750e+04,   2.15348125e+04,   1.97550469e+04,
         1.80409531e+04,   1.63841406e+04,   1.47842969e+04,
         1.32438281e+04,   1.17663438e+04,   1.03563906e+04,
         9.01753125e+03,   7.77323438e+03,   6.64531250e+03,
         5.62957812e+03,   4.71807812e+03,   3.90075000e+03,
         3.17331250e+03,   2.53498438e+03,   1.98378125e+03,
         1.51645312e+03,   1.12650000e+03,   8.08703125e+02,
         5.59656250e+02,   3.72718750e+02,   2.38671875e+02,
         1.46328125e+02,   8.44218750e+01,   4.56250000e+01,
         2.22500000e+01,   8.79687500e+00,   2.50000000e+00,
         5.15625000e-01])
array([ 301.09375 ,  300.53125 ,  300.015625,  299.4375  ,  298.765625,
        298.      ,  297.140625,  296.1875  ,  295.125   ,  293.953125,
        292.703125,  291.3125  ,  289.921875,  288.4375  ,  287.078125,
        285.5625  ,  284.078125,  282.453125,  280.578125,  279.15625 ,
        277.203125,  274.859375,  272.515625,  270.625   ,  269.5     ,
        268.96875 ,  266.875   ,  263.75    ,  262.6875  ,  259.71875 ,
        256.84375 ,  254.1875  ,  251.21875 ,  247.625   ,  243.890625,
        240.109375,  236.453125,  233.140625,  230.484375,  227.125   ,
        224.65625 ,  222.28125 ,  218.125   ,  213.25    ,  208.578125,
        203.90625 ,  199.796875,  195.921875,  192.28125 ,  195.171875,
        199.015625,  203.625   ,  207.703125,  210.421875,  214.25    ,
        218.203125,  223.078125,  227.640625,  229.859375,  232.703125,
        238.28125 ,  244.84375 ,  255.296875,  261.71875 ,  260.046875,
        271.171875,  243.546875,  208.515625,  170.84375 ,  223.03125 ])

%matplotlib notebook
import numpy as np
xs = np.linspace(0.1, 90000, 1000)[::-1]
print(press.data.shape)
ys_cubic = stratify.interpolate(xs, press.data[::3], temp.data[::3],
                          interpolation='cubic',
                          extrapolation='nan')

ys_linear = stratify.interpolate(xs, press.data[::3], temp.data[::3],
                          interpolation='linear',
                          extrapolation='nan')
plt.plot(ys_cubic, xs, label='cubic')
plt.plot(ys_linear, xs, label='linear')
plt.legend()
plt.gca().invert_yaxis()
plt.show()

download

I'm going to close this until future development is required...

@pelson pelson closed this Jan 31, 2017
@pelson pelson reopened this Feb 1, 2017
stickler-ci
stickler-ci reviewed Feb 1, 2017
View reviewed changes
stratify/tests/performance.py
interp_and_extrap(shape=(500, 600, 100))
parser = argparse.ArgumentParser(description='Run an interpolation.')
parser.add_argument('--shape', default='500,600,100',
help='The shape of the array to interpolate. Comma separated (no spaces).')
Copy link

@stickler-ci stickler-ci Feb 1, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

E501 line too long (99 > 79 characters)

stickler-ci
stickler-ci reviewed Feb 1, 2017
View reviewed changes
stratify/tests/test_vinterp.py

def test_inbetween(self):
x = np.linspace(0, 4, 5)
xs = [0.5, 1.25, 2.5, 3.75]
Copy link

@stickler-ci stickler-ci Feb 1, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

W291 trailing whitespace

@pelson pelson closed this Feb 1, 2017
@pelson pelson mentioned this pull request Feb 1, 2017
Closed
@cpelley
Copy link

cpelley commented Oct 31, 2017

Hey @pelson just noticed this. Really cool, is this interest driven? I see you closed it.
If you decide to get this back up, I'll happily take a look for you.

@pelson
Copy link
Member Author

pelson commented Nov 1, 2017

I didn't delete the branch for a reason 😄 .

Seems like an obvious extension to stratify, and one I'd love to see in here. Currently nobody actually needs this though (linear is typically good enough, even with atmospheric pressure, it seems)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

1 more reviewer

@stickler-ci stickler-ci stickler-ci left review comments

Reviewers whose approvals may not affect merge requirements

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@pelson @cpelley @stickler-ci