Dedensification

Examples of dedensification of a graph. Dedensification retains the structural pattern of the original graph and will only add compressor nodes when doing so would result in fewer edges in the compressed graph.

import matplotlib.pyplot as plt
import networkx as nx

plt.suptitle("Dedensification")

original_graph = nx.DiGraph()
white_nodes = ["1", "2", "3", "4", "5", "6"]
red_nodes = ["A", "B", "C"]
node_sizes = [250 for node in white_nodes + red_nodes]
node_colors = ["white" for n in white_nodes] + ["red" for n in red_nodes]

original_graph.add_nodes_from(white_nodes + red_nodes)
original_graph.add_edges_from(
    [
        ("1", "C"),
        ("1", "B"),
        ("2", "C"),
        ("2", "B"),
        ("2", "A"),
        ("3", "B"),
        ("3", "A"),
        ("3", "6"),
        ("4", "C"),
        ("4", "B"),
        ("4", "A"),
        ("5", "B"),
        ("5", "A"),
        ("6", "5"),
        ("A", "6"),
    ]
)
base_options = {"with_labels": True, "edgecolors": "black"}
pos = {
    "3": (0, 1),
    "2": (0, 2),
    "1": (0, 3),
    "6": (1, 0),
    "A": (1, 1),
    "B": (1, 2),
    "C": (1, 3),
    "4": (2, 3),
    "5": (2, 1),
}
ax1 = plt.subplot(1, 2, 1)
plt.title("Original (%s edges)" % original_graph.number_of_edges())
nx.draw_networkx(original_graph, pos=pos, node_color=node_colors, **base_options)

nonexp_graph, compression_nodes = nx.summarization.dedensify(
    original_graph, threshold=2, copy=False
)
nonexp_node_colors = list(node_colors)
nonexp_node_sizes = list(node_sizes)
for node in compression_nodes:
    nonexp_node_colors.append("yellow")
    nonexp_node_sizes.append(600)
plt.subplot(1, 2, 2)

plt.title("Dedensified (%s edges)" % nonexp_graph.number_of_edges())
nonexp_pos = {
    "5": (0, 0),
    "B": (0, 2),
    "1": (0, 3),
    "6": (1, 0.75),
    "3": (1.5, 1.5),
    "A": (2, 0),
    "C": (2, 3),
    "4": (3, 1.5),
    "2": (3, 2.5),
}
c_nodes = list(compression_nodes)
c_nodes.sort()
for spot, node in enumerate(c_nodes):
    nonexp_pos[node] = (2, spot + 2)
nx.draw_networkx(
    nonexp_graph,
    pos=nonexp_pos,
    node_color=nonexp_node_colors,
    node_size=nonexp_node_sizes,
    **base_options,
)

plt.tight_layout()
plt.show()