How to solve the warning "MatplotlibDeprecationWarning" and improve the plotting result

Community
1

Hi
This is my code

import numpy as np; 
import matplotlib; import matplotlib.pyplot as plt

frequency = np.array([0.0000,0.0125,0.0750,0.1250,0.2250,0.2000,0.2125,0.1000,0.0375,0.0125])
value_scale = np.arange(0, len(frequency), 1)

mu = 5; sigma = 2; params = ([mu, sigma])
gaussian = 1/(sigma*np.sqrt(2*np.pi)) * np.exp(-(value_scale-mu)**2/(2*sigma**2))

Module_square = (frequency - gaussian)**2
SSR = np.sum(Module_square)

mu_set = np.arange(0, 10, 0.01); sigma_set = np.arange(0.5, 5.5, 0.01)
mu_scale = mu_set; sigma_scale = np.matrix(sigma_set); value_mesh_scale = np.matrix(value_scale)
gaussian_mu = np.empty((len(sigma_scale), len(frequency)))
cycle = 0

for mu_var in mu_scale:
    gaussian_mu = np.exp(1/2*np.square(sigma_scale.T)*(-np.square(value_mesh_scale-mu_var)))/(sigma_scale.T*np.sqrt(2*np.pi))
    SSR_component = np.sum(np.matrix(np.square(frequency - gaussian_mu)), axis=1); cycle = cycle + 1
    if (cycle == 1):
        SSR_mesh = SSR_component
    else:
        SSR_mesh = np.hstack((SSR_mesh,SSR_component))

plt.figure(dpi=128); params = {"text.usetex":True, "font.family":"serif", "mathtext.fontset":"cm", "axes.titlesize": 16, "axes.labelsize":14, "figure.facecolor":"w"}
matplotlib.rcParams.update(params)

plt.pcolormesh(mu_scale, sigma_set, SSR_mesh, cmap="Blues_r", shading='gouraud')
plt.colorbar()

plt.show()

Screenshot_20220522_032809A
Screenshot_20220522_032809A3008×1324 210 KB

However, it notices me that

/tmp/ipykernel_24528/712004016.py:6: MatplotlibDeprecationWarning: For X (1001) and Y (501) with flat shading, the expected shape of A is (500, 1000). Passing A ((1, 500000)) is deprecated since 3.5 and will become an error two minor releases later.
  plt.pcolormesh(mu_scale, sigma_set, SSR_mesh, cmap="Blues_r", shading='auto')

I don’t want this warning. How to solve it?
Another question is that I want a logarithmical distributed colormap. How to realize it?

Thanks

2

You should probably use mu_mesh, sigma_mesh = np.meshgrid(mu_scale, sigma_scale) and work with those instead of the for loop to calculate SSR_mesh. For a logarithmic scale you could just replace SSR_mesh with np.log10(SSR_mesh) inside pcolormesh.

3

Thanks for your advice,

I solved this issue by the following:

mu_scale = np.arange(0, 10, 0.01); sigma_scale = np.arange(0.5, 5.5, 0.01); cycle = 0

mu_mesh, sigma_mesh = np.meshgrid(mu_scale, sigma_scale)

for value_var in value_scale:
    SSR_component = np.square(frequency[value_var]-np.exp(-(value_var-mu_mesh)**2/(2*sigma_mesh**2))/(sigma_mesh*np.sqrt(2*np.pi)))
    cycle = cycle + 1
    if (cycle == 1):
        SSR_mesh = SSR_component
    else:
        SSR_mesh = SSR_mesh + SSR_component

For the graph ploting, what I mean is that make the color map in logarithmic coordinates. I don’t want to change the data calculated.

4

Please see matplotlib.colors.LogNorm — Matplotlib 3.6.0 documentation and Colormap Normalizations — Matplotlib 3.6.0 documentation for an overview

5

FWIW, you could replace the whole for loop with this:

SSR_mesh = sum(
    np.square(
        f
        - np.exp(-((i - mu_mesh) ** 2) / (2 * sigma_mesh**2))
        / (sigma_mesh * np.sqrt(2 * np.pi))
    )
    for i, f in enumerate(frequency)
)