In Depth 1 - Drawing nodes

Here we show the fuctionalities and parameters of xgi.draw_nodes(). It is similar to the networkx corresponding function (+ some bonus) and heavily relies on matplotlib’s scatter function.

import matplotlib.pyplot as plt
import numpy as np

import xgi

Les us first create a small toy hypergraph containing edges of different sizes.

edges = [[1, 2, 3], [3, 4, 5], [3, 6], [6, 7, 8, 9], [1, 4, 10, 11, 12], [1, 4]]

H = xgi.Hypergraph(edges)

pos = xgi.barycenter_spring_layout(H, seed=42)  # fix position

xgi.draw(H, pos);

Basics

Let’s jump right into how xgi.draw_nodes() works. By default, it gives:

ax, node_collection = xgi.draw_nodes(H, pos)

Note that it returns a tuple (Axes, PathCollection). The PathCollection is what matplotlib’s plt.scatter() returns and can be used later to plot a colorbar.

The color, size, linewidth, and shape of the nodes can all be customised:

ax, node_collection = xgi.draw_nodes(
    H, pos, node_fc="b", node_ec="r", node_shape="v", node_size=25, node_lw=3
)

Colormaps

In XGI, you can easily color nodes according to a NodeStat, or just an array or a dict:

ax, node_collection = xgi.draw_nodes(
    H, pos, node_fc=H.nodes.degree(), node_labels=True, node_size=15
)
print(H.nodes.degree().asdict())

{1: 3, 2: 1, 3: 3, 4: 3, 5: 1, 6: 2, 7: 1, 8: 1, 9: 1, 10: 1, 11: 1, 12: 1}

By default, the colormap used is "Reds". To visualise the values corresponding to the colors, nothing easier than plotting a colorbar:

ax, node_collection = xgi.draw_nodes(
    H, pos, node_fc=H.nodes.degree(), node_labels=True, node_size=15
)
print(H.nodes.degree().asdict())
plt.colorbar(node_collection)
plt.show()

{1: 3, 2: 1, 3: 3, 4: 3, 5: 1, 6: 2, 7: 1, 8: 1, 9: 1, 10: 1, 11: 1, 12: 1}

You can specify a vmin and vmax for the node colors:

ax, node_collection = xgi.draw_nodes(H, pos, node_fc=H.nodes.degree(), vmin=2)
plt.colorbar(node_collection)
plt.show()

It’s also easy to specify a different colormap:

ax, node_collection = xgi.draw_nodes(
    H, pos, node_fc=H.nodes.degree(), node_fc_cmap="Greens"
)

plt.colorbar(node_collection)
plt.show()

And of course, all the colorbar customisation offered by matplotlib are available, for example:

ax, node_collection = xgi.draw_nodes(
    H, pos, node_fc=H.nodes.degree(), node_fc_cmap="Greens"
)

plt.colorbar(node_collection, label="Node Degree", shrink=0.8, ticks=[1, 2, 3])

<matplotlib.colorbar.Colorbar at 0x17e8018b0>

Combine with edges

This can be either via xgi.draw():

ax, collections = xgi.draw(H, pos=pos, node_fc=H.nodes.degree())

(node_collection, _, _) = collections

plt.colorbar(node_collection, label="Node Degree", shrink=0.8, ticks=[1, 2, 3])

<matplotlib.colorbar.Colorbar at 0x17e9d85e0>

Or by combining it with xgi.hyperedges() for more control:

ax, node_collection = xgi.draw_nodes(H, pos=pos, node_fc=H.nodes.degree(), zorder=5)

ax = xgi.draw_hyperedges(H, pos=pos)

plt.colorbar(node_collection, label="Node Degree", shrink=0.8, ticks=[1, 2, 3])

<matplotlib.colorbar.Colorbar at 0x17ea773d0>

Notice that here we had to specify the zorder of the nodes which is 0 by default in xgi.draw_nodes()!

Non-finite values for node colors

By default, non-finite values (inf, nan) are plotted in the “bad” color of the cmap. Here, it is white by default:

values = [np.inf, np.nan] + list(range(H.num_nodes - 2))  # values for node colors

ax, node_collection = xgi.draw_nodes(
    H, pos, node_fc=values, node_fc_cmap="viridis", node_labels=True, node_size=15
)

plt.colorbar(node_collection, label="Node Degree", shrink=0.8)
plt.show()

The default can be changed, for example here to red:

values = [np.inf, np.nan] + list(range(H.num_nodes - 2))

cmap = plt.cm.viridis
cmap.set_bad("red")  # set bad

ax, node_collection = xgi.draw_nodes(
    H, pos, node_fc=values, node_fc_cmap=cmap, node_labels=True, node_size=15
)

plt.colorbar(node_collection, label="Node Degree", shrink=0.8)
plt.show()