CODEWITHSUNDEEP
pythonscipyinterpolation
maximusdooku
maximusdooku

Reputation: 5512

Why can I not interpolate even though the values are monotonically increasing?

I have 2 arrays that I plot this way:

plt.plot(x1, x2)

enter image description here

And I want to find out what are the values of x2 at x1=2,5,75,10,100, and 1000. Since all values may not exist, I interpolate using this:

from scipy.interpolate import interp1d
f1=interp1d(x1, x2)
f2=interp1d(x1, x2, kind='cubic')

But f2 throws this error:

ValueError: Expect x to be a 1-D sorted array_like.

But the values are monotonically increasingly as expected by: https://docs.scipy.org/doc/scipy/reference/generated/scipy.interpolate.interp1d.html

What am I doing wrong?

x1

array([1.00020004e+00, 1.00020004e+00, 1.00080064e+00, 1.00200401e+00,
       1.00341160e+00, 1.00603622e+00, 1.01173614e+00, 1.02165917e+00,
       1.02965404e+00, 1.04362346e+00, 1.05820106e+00, 1.07688994e+00,
       1.10107906e+00, 1.12688754e+00, 1.16063138e+00, 1.19161106e+00,
       1.23578843e+00, 1.29198966e+00, 1.35062129e+00, 1.42775557e+00,
       1.52207002e+00, 1.63345312e+00, 1.77746178e+00, 1.93124759e+00,
       2.11954218e+00, 2.37191651e+00, 2.68528464e+00, 2.97973778e+00,
       3.38983051e+00, 3.89105058e+00, 4.48430493e+00, 5.31349628e+00,
       6.28930818e+00, 7.59878419e+00, 9.38086304e+00, 1.20192308e+01,
       1.57232704e+01, 2.06611570e+01, 2.68817204e+01, 3.49650350e+01,
       4.58715596e+01, 6.57894737e+01, 8.92857143e+01, 1.38888889e+02,
       2.27272727e+02, 4.16666667e+02, 1.00000000e+03, 2.50000000e+03,
       2.50000000e+03, 5.00000000e+03])

x2

array([ 11.18083778,  12.00065196,  12.88057749,  13.82502193,
        14.838716  ,  15.92673731,  17.09453577,  18.34796088,
        19.69329106,  21.13726508,  22.68711581,  24.35060647,
        26.1360695 ,  28.05244828,  30.10934199,  32.31705361,
        34.68664159,  37.22997521,  39.95979405,  42.88977178,
        46.0345847 ,  49.40998519,  53.03288065,  56.9214182 ,
        61.09507555,  65.57475857,  70.38290604,  75.54360201,
        81.08269642,  87.02793465,  93.40909644, 100.25814508,
       107.6093875 , 115.49964612, 123.96844331, 133.05819935,
       142.81444487, 153.28604899, 164.52546404, 176.58898835,
       189.53704818, 203.43450047, 218.35095766, 234.36113639,
       251.54523176, 269.98931903, 289.78578477, 311.03378962,
       333.83976499, 358.31794618])

Upvotes: 0

Views: 5783

Answers (2)

Mark
Mark

Reputation: 92440

The test it's doing it is:

np.any(x1[1:] <= x1[:-1])

notice the <=, which means equal values (like: 2.50000000e+03,, 2.50000000e+03) in your array will cause errors.

Changing these and (there's two more at the start of the array) will make the error go away.

I not sure this is the right way to go about dealing with your data, but you could delete the dupes (and corresponding x2 values) with something like:

x3, ind = np.unique(x1, return_index = True)
x4 = x2[ind] 
f2=interp1d(x3, x4, kind='cubic')

Upvotes: 1

user2261062
user2261062

Reputation:

Just add this two lines before doing the interpolation:

x1[0] -= 0.0000001
x1[-2] += 0.0000001

Looks like with epsilon does not work

Upvotes: 1

Related Questions