Interpolation in SciPy: Finding X that produces Y
-
06-07-2019 - |
Question
Is there a better way to find which X gives me the Y I am looking for in SciPy? I just began using SciPy and I am not too familiar with each function.
import numpy as np
import matplotlib.pyplot as plt
from scipy import interpolate
x = [70, 80, 90, 100, 110]
y = [49.7, 80.6, 122.5, 153.8, 163.0]
tck = interpolate.splrep(x,y,s=0)
xnew = np.arange(70,111,1)
ynew = interpolate.splev(xnew,tck,der=0)
plt.plot(x,y,'x',xnew,ynew)
plt.show()
t,c,k=tck
yToFind = 140
print interpolate.sproot((t,c-yToFind,k)) #Lowers the spline at the abscissa
Solution
The UnivariateSpline class in scipy makes doing splines much more pythonic.
x = [70, 80, 90, 100, 110]
y = [49.7, 80.6, 122.5, 153.8, 163.0]
f = interpolate.UnivariateSpline(x, y, s=0)
xnew = np.arange(70,111,1)
plt.plot(x,y,'x',xnew,f(xnew))
To find x at y then do:
yToFind = 140
yreduced = np.array(y) - yToFind
freduced = interpolate.UnivariateSpline(x, yreduced, s=0)
freduced.roots()
I thought interpolating x in terms of y might work but it takes a somewhat different route. It might be closer with more points.
OTHER TIPS
If all you need is linear interpolation, you could use the interp function in numpy.
I may have misunderstood your question, if so I'm sorry. I don't think you need to use SciPy. NumPy has a least squares function.
#!/usr/bin/env python
from numpy.linalg.linalg import lstsq
def find_coefficients(data, exponents):
X = tuple((tuple((pow(x,p) for p in exponents)) for (x,y) in data))
y = tuple(((y) for (x,y) in data))
x, resids, rank, s = lstsq(X,y)
return x
if __name__ == "__main__":
data = tuple((
(1.47, 52.21),
(1.50, 53.12),
(1.52, 54.48),
(1.55, 55.84),
(1.57, 57.20),
(1.60, 58.57),
(1.63, 59.93),
(1.65, 61.29),
(1.68, 63.11),
(1.70, 64.47),
(1.73, 66.28),
(1.75, 68.10),
(1.78, 69.92),
(1.80, 72.19),
(1.83, 74.46)
))
print find_coefficients(data, range(3))
This will return [ 128.81280358 -143.16202286 61.96032544].
>>> x=1.47 # the first of the input data
>>> 128.81280358 + -143.16202286*x + 61.96032544*(x**2)
52.254697219095988
0.04 out, not bad
Licensed under: CC-BY-SA with attribution
Not affiliated with StackOverflow