0

Let's say I have the following matrix (as a numpy.matrix or np.array object):

#x/column labels: = -1, 0, 1

#y/row labels: 0, 1, 2 (like normal)

data: np.matrix([[1, 2, 3], [-1, np.nan, -2], [7, -5, -6]])

plt.matshow(data, colorbar = 'RdBu')

plt.colorbar()

plt.show()

I want to use plt.matshow() on this matrix. First, I don't know why, but sometimes, my np.matrix objects display like this if I simply print them out (just ignore that the numbers are slightly different, please):

matrix([[ 1.,  2.,  3.],
    [-1., nan, -2.],
    [-3.,  5.,  2.]])

These will be accepted by np.matshow(). However sometimes they appear like this:

matrix([[1, 2, 3],
    [-1, nan, -2],
    [7, -5, -6]], dtype=object)

First question:

If that dtype=object is there, np.matshow() will not plot them if a nan or None is present. I don't know why this tag is there sometimes and sometimes not. I generate these through a somewhat convoluted script, and I suspect it may be sometimes because I am doing np.matrix(a) where a is some python array or np.array. Could that be the cause? I have played around with simple cases in terminal windows but have not been able to figure it out. Sometimes the "error marker" is np.nan, sometimes it's python's None.

Second question: When there is a np.nan in the matrix, matshow() just paints that square white. This is okay if the color map used, like the default one, does not include white. However, I want to use a "binary" type color scale like RdBu; positive for blue, negative for red, white for 0. Is there a way to, for example, put a big green X over nan's, or make its square black, etc? Basically, a way to clearly distinguish nan's from 0's in this color map.

Third question: please notice in my first code snipped that the x/column labels are supposed to be -1, 0, 1. Since the np.matrix object seems to not have custom indices, it just plots starting from 0. Is there a way to pass this info to np.matshow? If I use plt.xlim([-1, 1]) for example, it just only shows two of the rows of the matrix, since there is no data at -1, and the third row is at x = 2.

Sheldore
  • 37,862
  • 7
  • 57
  • 71
iammax
  • 453
  • 2
  • 5
  • 18

2 Answers2

1

You can first make your return a numpy float array like this:

data = np.array(data).astype(np.float)
ImportanceOfBeingErnest
  • 321,279
  • 53
  • 665
  • 712
1

Partial solution to your question: Similar problem (the colormap thing) has been partially addressed here: How can I plot NaN values as a special color with imshow in matplotlib? ... Specifically for your array, it would be as follows where you can choose any color nan_color you like to plot the nan values. 

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

data =  np.matrix([[1, 2, 3], [-1, np.nan, -2], [7, -5, -6]])
data_masked = np.ma.array(data, mask=np.isnan(data))
nan_color = 'black' # or 'white' or whatever you like
color = cm.RdBu
color.set_bad(nan_color,1.)
plt.matshow(data_masked, cmap=color)
plt.colorbar()
Sheldore
  • 37,862
  • 7
  • 57
  • 71