Disparity Filter Visualization

This example applies the disparity filter to nx.les_miserables_graph() using a strict score-then-filter workflow:

1. Score all edges on the full graph with disparity_filter.

2. Filter by disparity_pvalue with threshold_filter.

3. Visualize removed and retained structure with compare_graphs, in the colors of the documentation’s figure style (docs/_figures/style.py).

Original edges: 254
Backbone edges: 9
Removed nodes: 67
Removed edges: 245

import tempfile
import warnings
from pathlib import Path

import networkx as nx
from _figures import style

import nx_backbone as nb
from nx_backbone.visualization import (
    compare_graphs,
    graph_difference,
    save_graph_comparison,
)

G = nx.les_miserables_graph()

# Score on the full graph.
scored = nb.disparity_filter(G)

# Then filter scored edges to extract the backbone.
backbone = nb.threshold_filter(
    scored,
    score="disparity_pvalue",
    threshold=0.05,
    mode="below",
    include_all_nodes=False,
)

if backbone.number_of_edges() == G.number_of_edges():
    warnings.warn(
        "Disparity filter returned the same number of edges as the original "
        "graph. Re-test and validate threshold settings.",
        UserWarning,
        stacklevel=1,
    )

pos = style.layout(G, "les_miserables")  # the layout every docs figure uses
colors = style.palette()
fig, ax, diff = compare_graphs(
    G,
    backbone,
    pos=pos,
    node_size=40,
    kept_node_color=colors.node,
    removed_node_color=colors.node_removed,
    base_edge_color=colors.removed,
    kept_edge_color=colors.kept,
    base_edge_width=0.6,
    kept_edge_width=1.5,
    title="disparity_filter on Les Misérables (disparity_pvalue < 0.05)",
    return_diff=True,
)

diff_summary = graph_difference(G, backbone)
_ = save_graph_comparison(
    G,
    backbone,
    output_path=Path(tempfile.gettempdir()) / "disparity_filter_example.png",
    pos=pos,
    title="disparity_filter on Les Misérables",
)

print(f"Original edges: {G.number_of_edges()}")
print(f"Backbone edges: {len(diff['kept_edges'])}")
print(f"Removed nodes: {len(diff['removed_nodes'])}")
print(f"Removed edges: {len(diff_summary['removed_edges'])}")