How to plot a sparse matrix in a scatter pairs plot in python while discarding zero values?

Question

I have a sparse matrix (140 x 363) with many zeroes. This matrix only has positive float values.

I would like to generate a scatter pairs matrix plot discarding all zero values (0.0, 0.0) without affecting the display of the example values (0.0, 7.2), (7.2, 0.0), etc.

I have investigated sns.pairplot, sns.pairgrid, pd.scatter_matrix, Altair, and Plotty without getting a solution.

Unfortunately, none of these tools accept crs, coo, etc.

I have also tried plotting with the full matrix by passing nonzero instructions to the dataframe but it doesn't work either.

Is there a solution?

Answer 1

Can't you convert your sparse matrix to a coo format and then take the x and y coordinates?

import numpy as np
import pandas as pd
import seaborn as sns
from scipy.sparse import csr_matrix, coo_matrix

# demo data
row = np.array([0, 0, 1, 2, 2, 2])
col = np.array([0, 2, 2, 0, 1, 2])
data = np.array([1.1, 2.2, 0, 4.4, 5.5, 6.6])

# sparse matrix
sparse_matrix = csr_matrix((data, (row, col)), shape=(3, 3))

# to coo
sparse_matrix = sparse_matrix.tocoo()

# remove point which value are zero
sparse_matrix.eliminate_zeros()

# convert to pandas
df = pd.DataFrame({'x': sparse_matrix.row,
                   'y': sparse_matrix.col,
                   'value': sparse_matrix.data})

# plot
sns.scatterplot(data=df, x="x", y="y")