Run in Google Colab

View on GitHub

Visualizing networks¶

Let's showcase a few ways you can visualize networks using the igraph library. Let's first import a few libraries:

In [1]:

import igraph as ig
import numpy as np

import igraph as ig import numpy as np

Now let's load one of the networks in the Datasets/Networks folder. Let's load the EUPowerGrid network, which is a network of the European power grid.

We are going to use the simplify() function to remove self-loops and multiple edges, which is a common preprocessing step for network analysis.

In [2]:

# Loading EUPowerGrid network
g_power = ig.Graph.Read_GML("../../Datasets/Networks/EUPowerGrid.gml").simplify() # adjust the path as needed.

# Loading EUPowerGrid network g_power = ig.Graph.Read_GML("../../Datasets/Networks/EUPowerGrid.gml").simplify() # adjust the path as needed.

Let's see the attributes of the network:

In [3]:

# Node attributes
print(g_power.vertex_attributes())

# Node attributes print(g_power.vertex_attributes())

['id', 'name', 'Country']

We can calculate the layout using the layout function. Let's start with the circle layout with the order defined by the degree of the nodes.

Note that the default plot function from igraph needs you to install cairo. However this library can be a bit tricky to install on some systems. So I will offer here an alternative way to plot the network using matplotlib.

In [7]:

# Size of node changes with degree
node_degrees = np.array(g_power.degree())
layout = g_power.layout("circle",order=node_degrees.argsort())
# Plotting using the cairo backend (requires cairo installed)
# ig.plot(g_power,
#     layout=layout, 
#     vertex_size=node_degrees,
#     )

# Size of node changes with degree node_degrees = np.array(g_power.degree()) layout = g_power.layout("circle",order=node_degrees.argsort()) # Plotting using the cairo backend (requires cairo installed) # ig.plot(g_power, # layout=layout, # vertex_size=node_degrees, # )

Here is a nice matplotlib function to plot the network.

In [8]:

import matplotlib.pyplot as plt
import numpy as np

def plot_igraph_with_matplotlib(
    g,
    layout,
    node_size=300,
    node_color="skyblue",
    node_alpha=1.0,
    edge_color="gray",
    edge_width=1.0,
    edge_alpha=0.7,
    with_labels=False,
    label_attr="label",
    label_font_size=10,
    label_color="black",
    figsize=(6, 6),
    dpi=100,
    axis_off=True
):
    """
    Draw an igraph Graph `g` using matplotlib, given `layout`.
    
    Parameters
    ----------
    g : igraph.Graph
      The graph to draw.
    layout : Layout or sequence of (x, y)
      An igraph Layout object or list/array of coordinate pairs.
    node_size : float
      Size of the nodes (passed to plt.scatter `s`).
    node_color : color or list of colors
      Node face color.
    node_alpha : float
      Node transparency (0.0 transparent, 1.0 opaque).
    edge_color : color
      Color for all edges.
    edge_width : float
      Line width for edges.
    edge_alpha : float
      Edge transparency.
    with_labels : bool
      Whether to draw vertex labels.
    label_attr : str
      Vertex attribute name to use for labels; if absent, vertex indices used.
    label_font_size : float
      Font size for labels.
    label_color : color
      Color for label text.
    figsize : tuple
      Matplotlib figure size.
    dpi : int
      Figure DPI.
    axis_off : bool
      If True, hides axes.
    """
    # extract coordinates as an (N,2) array
    coords = np.array(layout.coords) if hasattr(layout, "coords") else np.array(layout)
    xs, ys = coords[:, 0], coords[:, 1]

    fig, ax = plt.subplots(figsize=figsize, dpi=dpi)

    # draw edges
    for src, tgt in g.get_edgelist():
        x0, y0 = coords[src]
        x1, y1 = coords[tgt]
        ax.plot(
            [x0, x1],
            [y0, y1],
            color=edge_color,
            linewidth=edge_width,
            alpha=edge_alpha,
            zorder=1
        )

    # draw nodes
    ax.scatter(
        xs,
        ys,
        s=node_size,
        c=node_color,
        alpha=node_alpha,
        zorder=2,
        edgecolors="black"
    )

    # draw labels
    if with_labels:
        if label_attr in g.vs.attribute_names():
            labels = g.vs[label_attr]
        else:
            labels = [str(i) for i in range(g.vcount())]
        for idx, label in enumerate(labels):
            ax.text(
                xs[idx],
                ys[idx],
                label,
                fontsize=label_font_size,
                color=label_color,
                ha="center",
                va="center",
                zorder=3
            )

    if axis_off:
        ax.set_axis_off()

    plt.tight_layout()
    plt.show()

import matplotlib.pyplot as plt import numpy as np def plot_igraph_with_matplotlib( g, layout, node_size=300, node_color="skyblue", node_alpha=1.0, edge_color="gray", edge_width=1.0, edge_alpha=0.7, with_labels=False, label_attr="label", label_font_size=10, label_color="black", figsize=(6, 6), dpi=100, axis_off=True ): """ Draw an igraph Graph `g` using matplotlib, given `layout`. Parameters ---------- g : igraph.Graph The graph to draw. layout : Layout or sequence of (x, y) An igraph Layout object or list/array of coordinate pairs. node_size : float Size of the nodes (passed to plt.scatter `s`). node_color : color or list of colors Node face color. node_alpha : float Node transparency (0.0 transparent, 1.0 opaque). edge_color : color Color for all edges. edge_width : float Line width for edges. edge_alpha : float Edge transparency. with_labels : bool Whether to draw vertex labels. label_attr : str Vertex attribute name to use for labels; if absent, vertex indices used. label_font_size : float Font size for labels. label_color : color Color for label text. figsize : tuple Matplotlib figure size. dpi : int Figure DPI. axis_off : bool If True, hides axes. """ # extract coordinates as an (N,2) array coords = np.array(layout.coords) if hasattr(layout, "coords") else np.array(layout) xs, ys = coords[:, 0], coords[:, 1] fig, ax = plt.subplots(figsize=figsize, dpi=dpi) # draw edges for src, tgt in g.get_edgelist(): x0, y0 = coords[src] x1, y1 = coords[tgt] ax.plot( [x0, x1], [y0, y1], color=edge_color, linewidth=edge_width, alpha=edge_alpha, zorder=1 ) # draw nodes ax.scatter( xs, ys, s=node_size, c=node_color, alpha=node_alpha, zorder=2, edgecolors="black" ) # draw labels if with_labels: if label_attr in g.vs.attribute_names(): labels = g.vs[label_attr] else: labels = [str(i) for i in range(g.vcount())] for idx, label in enumerate(labels): ax.text( xs[idx], ys[idx], label, fontsize=label_font_size, color=label_color, ha="center", va="center", zorder=3 ) if axis_off: ax.set_axis_off() plt.tight_layout() plt.show()

In [11]:

# Size of node changes with degree
node_degrees = np.array(g_power.degree())
layout = g_power.layout("circle",order=node_degrees.argsort())
# using the custom plotting function
plot_igraph_with_matplotlib(
    g_power,
    layout=layout,
    node_size=node_degrees * 20,  # Scale node size by degree
    edge_alpha=0.05
)

# Size of node changes with degree node_degrees = np.array(g_power.degree()) layout = g_power.layout("circle",order=node_degrees.argsort()) # using the custom plotting function plot_igraph_with_matplotlib( g_power, layout=layout, node_size=node_degrees * 20, # Scale node size by degree edge_alpha=0.05 )

Let's try another layout, the Fruchterman-Reingold layout, which is a force-directed layout. This layout is often used for visualizing networks as it tends to spread out the nodes nicely. We can color the nodes based on the country they belong to.

In [18]:

# color countries
from collections import Counter
import matplotlib as mpl
country2Index = {country:index for index,(country,_) in enumerate(Counter(g_power.vs["Country"]).most_common(10))}
countryColors = [mpl.cm.tab10(country2Index[country]) if country in country2Index else "#888888" for country in g_power.vs["Country"]]
# Size of node changes with degree
node_degrees = np.array(g_power.degree())
layout = g_power.layout("fruchterman_reingold", niter= 2000)

# Other supported layouts: drl, davidson_harel, circle, kamada_kawai, fruchterman_reingold, graphopt, mds

# ig.plot(g_power,
#     layout=layout,
#     vertex_size=node_degrees,
#     vertex_color=countryColors
#     )

# custom plotting with matplotlib
plot_igraph_with_matplotlib(
    g_power,
    layout=layout,
    node_size=node_degrees * 20,  # Scale node size by degree
    node_color=countryColors,
    figsize=(10, 10)
)

# color countries from collections import Counter import matplotlib as mpl country2Index = {country:index for index,(country,_) in enumerate(Counter(g_power.vs["Country"]).most_common(10))} countryColors = [mpl.cm.tab10(country2Index[country]) if country in country2Index else "#888888" for country in g_power.vs["Country"]] # Size of node changes with degree node_degrees = np.array(g_power.degree()) layout = g_power.layout("fruchterman_reingold", niter= 2000) # Other supported layouts: drl, davidson_harel, circle, kamada_kawai, fruchterman_reingold, graphopt, mds # ig.plot(g_power, # layout=layout, # vertex_size=node_degrees, # vertex_color=countryColors # ) # custom plotting with matplotlib plot_igraph_with_matplotlib( g_power, layout=layout, node_size=node_degrees * 20, # Scale node size by degree node_color=countryColors, figsize=(10, 10) )

Using Helios-Web¶

You can try other layouts on igraph. However if you want to try more advanced visualizations, you can use Helios-Web. Helios-Web is a web-based visualization tool that allows you to visualize networks in a more interactive way. You can upload your network and try different layouts and styles.

First you need to save the network to the .xnet format or .gml format.

You may need to install the xnetwork package to save the network in the .xnet format. You can install it using pip:

pip install xnetwork

In [20]:

import xnetwork as xn
xn.save(g_power, "EUPowerGrid.xnet")

import xnetwork as xn xn.save(g_power, "EUPowerGrid.xnet")

Now open a browser window and drag and drop the .xnet file into the Helios-Web interface. From the bottom menus you can select different attributes.

go to: https://heliosweb.io/docs/example/?advanced&dark

We are still working on the Helios-Web interface, so there are no controllers for the layout yet. If you know javascript, you can use helios-web as a library to create your own visualizations.

More information on how to use Helios-Web can be found in the Helios-Web documentation.

Exercises¶

Find the best layout for the EU power grid network.

Hint: Check documentation for layouts in https://igraph.org/python/api/latest/igraph._igraph.GraphBase.html#layout_fruchterman_reingold

In [ ]:

Load a network from the Networks folder, explore it, and create a visualization

In [ ]: