Develop Reference

Scipy strange import behaviour

Scipy has an different way of handling submodules to Numpy, for example
import scipy as sp
import numpy as np
A = np.eye(4)
np.linalg.det(A)
sp.linalg.det(A)
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
AttributeError: module 'scipy' has no attribute 'linalg'
This is moderately annoying because of the asymmetry with respect to Numpy, but it is exactly the behaviour that the documentation describes. The proper usage according to the docs is
from scipy import linalg
import numpy as np
A = np.eye(4)
np.linalg.det(A)
linalg.det(A) # using Scipy
which works just fine.
Now, here's the weird thing
import scipy as sp
import numpy as np
from scipy.linalg import expm # extra line inserted into first example
A = np.eye(4)
np.linalg.det(A)
sp.linalg.det(A)
then the Numpy-style code works just fine. The extra line causes linalg to be added to the namespace sp, a side effect of the extra import.
I get the programming pattern I want but the third line is not easy to explain in example code.
QUESTION: why does Scipy do this? is there any more straightforward way to have it so that Scipy behaves more Numpyish?

In fact, you almost never need or to want import scipy as sp or anything like that.
There is almost nothing in the top level scipy namespace. All useful stuff is in subpackages (one exception is LowLevelCallable which is in the top-level namespace). This way, users are better off either importing from subpackages, from scipy.signal import detrend, or importing subpackages themselves (from scipy import signal; signal.detrend(...)).
As to the disparity with numpy, numpy is very much the opposite: a lot of useful stuff is in the top-level namespace, so you import it from there.
Unless you're using np.linalg, np.random, np.fft or np.testing, which are public-facing usable submodules.

Import error. Can't import kmeans_plusplus

I'm trying to replicate an example of a clustering model with scikit-learn:
import sklearn
sklearn.__version__
Returns:
'0.23.2'
And:
from sklearn.cluster import kmeans_plusplus
Returns the Error message:
ImportError: cannot import name 'kmeans_plusplus' from 'sklearn.cluster' (C:\Users\sddss\anaconda3\lib\site-packages\sklearn\cluster\__init__.py)

According to the documentation, kmeans_plusplus is
New in version 0.24.
so it is not available for the version 0.23.2 you are using.
Nevertheless, this should not be a real issue; the only difference between the "good old" K-Means already available in scikit-learn is the initialization of the cluster centers according to the kmeans++ algorithm; and this is already available in the standard KMeans. From the standard KMeans documentation regarding the init argument:
'k-means++' : selects initial cluster centers for k-mean clustering in a smart way to speed up convergence
So, what you need to do instead is simply to use the "vanilla" KMeans of scikit-learn with the argument init='kmeans++':
from sklearn.cluster import KMeans
kmeans = KMeans(n_clusters=n_clusters, init='kmeans++')

There is no kmeans_plusplus class or module for version 0.23.2. You need to import KMeans and set the init key word argument to kmeans++ to obtain the behaviour you want
from sklearn.cluster import KMeans
kmeans = KMeans(init='k-means++')

Why doesn't this scipy fourier integration function work?

When I try to integrate a periodic array with the scipy function sp.fftpack.diff(x,order=-1), it sometimes works and sometimes doesn't.
For example, when integrating x=sin(alpha) to obtain an array of the values of the integral when evaluated from 0 to discrete values up to 2*pi I get the expected result -cos(\alphas). However, when I use it to calculate the values of the integrals of x=sin(alpha)+cos(alpha)+1 in the same ranges I do not get the right answer, even when the function is periodic.
I do not understand how this function works. Does someone have an idea?
https://docs.scipy.org/doc/scipy/reference/generated/scipy.fftpack.diff.html
For example, with this code I obtain the results in the image,I am also comparing the results with the obtained by the trapezoidal rule, which does work when fixing the offset.enter image description here
import numpy as np
from scipy import fftpack as sp
from scipy import integrate as inte
import matplotlib.pyplot as plt
N=150
h=(2*np.pi)/N
x=np.arange(-np.pi,np.pi,h)
y=np.sin(x)+np.cos(x)+1
arrExact=-np.cos(x)+np.sin(x)+x
st=inte.cumtrapz(y,x,initial=0)-2.1
di=sp.diff(y, order=-1)-1
plt.plot(x,di,label='diff')
plt.plot(x,arrExact,label='Exact')
plt.plot(x,st,label='cumpTrapz')
plt.legend()
plt.show()
Edit: Well, reading again I realized scipy assumes x[0]=0, however I need to integrate spectrally arrays that do not satisfies this condition, How can I proceed?

DeprecationWarning rasied in Spyder for every Scipy function used

I'm coding in Spyder and the code runs, but every line that uses sp.___ raises a DeprecationWarning, e.g. DeprecationWarning: scipy.array is deprecated and will be removed in SciPy 2.0.0, use numpy.array instead.
Why is Spyder doing this and how do I allow me to use scipy without raising this error? Failing that, what can I do to suppress the error from popping up each time?
The code is like this:
import matplotlib.pyplot as plt,scipy as sp
import scipy.optimize as op
a=9.3779
x_in=sp.array([.095,.065,.09,.108,.125,.115,.040,.055,.055])
x=(x_in+14)
y_in=sp.array([.2,.6,.5,.4,.1,.3,-0.2,-0.4,0])
y=y_in+45
ax.plot(x_in,y_in,'ro')
plt.show()
This raises the error:
C:\Users\Shiva Pingle\Desktop\python\others\peaks.py:38: DeprecationWarning: scipy.array is deprecated and will be removed in SciPy 2.0.0, use numpy.array instead
x_in=sp.array([.095,.065,.09,.108,.125,.115,.040,.055,.055])
C:\Users\Shiva Pingle\Desktop\python\others\peaks.py:40: DeprecationWarning: scipy.array is deprecated and will be removed in SciPy 2.0.0, use numpy.array instead
y_in=sp.array([.2,.6,.5,.4,.1,.3,-0.2,-0.4,0])

Your solution in the comments will make you ignore all the deprecation warnings. This is not suggested.
You could instead import numpy as np and use the np.array().
Corrected code:
import matplotlib.pyplot as plt,scipy as sp
import scipy.optimize as op
import numpy as np # Added import of numpy
a=9.3779
x_in=np.array([.095,.065,.09,.108,.125,.115,.040,.055,.055]) # Changed sp to np
x=(x_in+14)
y_in=np.array([.2,.6,.5,.4,.1,.3,-0.2,-0.4,0]) # Changed sp to np
y=y_in+45
plt.plot(x_in,y_in,'ro') # Also changed the ax to plt
plt.show()

skimage - TypeError: peak_local_max() got an unexpected keyword argument 'num_peaks_per_label'

The following code gives me the error present in the title :
from skimage.feature import peak_local_max
local_maxi = peak_local_max(imd,labels=iml,
indices=False,num_peaks_per_label=2)
Where imd is a "distance transformed image" which was obtained with :
from scipy import ndimage
imd = ndimage.distance_transform_edt(im)
im is the input binary image that I would like to later on segment with the watershed function of scikit-image. But to use this function properly, I first need to find the markers which will serve as the starting flooding points : that's what I'm trying to do with the 'peak_local_max' function.
Also, iml is the labeled version of im, that I got with :
from skimage.measure import label
iml = label(im)
I don't know what I've been doing wrong. Also, I've noticed that, the function seems to totally ignore its num_peaks argument. For instance, when I do :
local_maxi = peak_local_max(imd,labels=iml,
indices=True,num_peaks=1)
I always get the same number of peaks detected as when I set num_peaks=500 or num_peaks=np.inf. What am I missing here please ?

As #a_guest pointed out, my version of skimage wasn't matching with the version of the documentation I was referring to. The num_peaks_per_label argument is currently only available in the v0.13dev version. Updating my version to the dev version also fixed my problem with the num_peaks argument.