How to get a colorbar in networkx.draw_networkx?

Question

I am performing clustering on relational graphs, and am currently using networkx to visualize the result:

G = nx.from_numpy_matrix(X)
layout = nx.fruchterman_reingold_layout(G)
nx.draw_networkx(G, pos=layout, with_labels=True, node_color=predict, cmap=plt.cm.coolwarm, vmin=0, vmax=1)

Is it possible to get a colorbar? Simply using plt.colorbar() gives the error:

RuntimeError: No mappable was found to use for colorbar creation. First define a mappable such as an image (with imshow) or a contour set (with contourf).

And there aren't any options in the documentation. I am also open to use another package for visualization as long as it is compatible with Python 3.

It's not apparent how you create this graph from the question. What the error tells you is that you need to supply an argument to the colorbar function. You may use a [`matplotlib.cm.ScalarMappable`](https://matplotlib.org/api/cm_api.html#matplotlib.cm.ScalarMappable) which uses the same vmin, vmax and colormap as the networks graph and supply it as argument to the colorbar function. — ImportanceOfBeingErnest, Mar 30 '18 at 10:21
Updated the question. I am not sure how to use `matplotlib.cm.ScalarMappable` simply supplying it to `plt.colorbar()` didn't work. Any pointers? — Toke Faurby, Mar 30 '18 at 10:34
Well, according to the usual approach for Stackoverflow questions and answers, you may provide a [mcve] of the issue that I can run, modify and test, such that I can subsequently provide you with an answer, showing the exact code to use here. — ImportanceOfBeingErnest, Mar 30 '18 at 10:36
@TokeFaurby, can you provide values for `predict`, so we could reproduce it ? — MaxU - stand with Ukraine, Mar 30 '18 at 15:24

MaxU - stand with Ukraine · Accepted Answer · 2018-03-30T16:18:47.053

Here is a demo based on Karate Club graph:

import pandas as pd
import networkx as nx    
import matplotlib.pyplot as plt

G = nx.karate_club_graph()

df = (pd.DataFrame(list(G.degree), columns=['node','degree'])
        .set_index('node'))
df['club'] = pd.Series({node:data['club']
                        for node,data in G.nodes(data=True)})
df['color'] = df.groupby('club')['degree'].transform(lambda c: c/c.max())
df.loc[df['club']=='Officer', 'color'] *= -1

layout = nx.fruchterman_reingold_layout(G)
vmin = df['color'].min()
vmax = df['color'].max()
cmap = plt.cm.coolwarm

nx.draw_networkx(G, pos=layout, with_labels=True, node_color=df['color'],
                 cmap=cmap, vmin=vmin, vmax=vmax)
sm = plt.cm.ScalarMappable(cmap=cmap, norm=plt.Normalize(vmin=vmin, vmax=vmax))
sm.set_array([])
cbar = plt.colorbar(sm)

Result:

How to get a colorbar in networkx.draw_networkx?

1 Answers1