nx-backbone

Backbone extraction algorithms for complex networks, built on NetworkX.

nx-backbone provides about 110 functions across 14 modules for extracting backbone structures from weighted, unweighted, bipartite, temporal, and higher-order (hypergraph) networks. The public API is type-annotated, and the graph-in, graph-out functions dispatch to NetworkX backends (see Backends). Core usage follows a score-then-filter pattern: methods score edges on a full graph first, then filtering utilities extract the final sparse backbone.

Tip

New here? After the Quick Start, start with From a hairball to a backbone, end to end, a worked case study of the whole workflow: why a backbone, scoring versus filtering, choosing alpha and corrections, comparing and validating backbones, and reporting them.

Module	Description	Key Functions
Statistical Methods	Hypothesis-testing methods	disparity_filter(), noise_corrected_filter()
Structural Methods	Topology-based methods	high_salience_skeleton(), metric_backbone()
Proximity Methods	Neighborhood-similarity scoring	jaccard_backbone(), cosine_backbone()
Hybrid Methods	Combined approaches	glab_filter()
Bipartite Methods	Bipartite projection backbones	sdsm(), fdsm(), fixedrow()
Hypergraph Methods	Hypergraph backbones and interoperability	mdl_hypergraph_backbone(), statistically_validated_hypergraph(), read_hif()
Unweighted Sparsification	Sparsification for unweighted graphs	sparsify(), lspar()
Filters	Post-hoc filtering utilities	threshold_filter(), fraction_filter()
Measures	Evaluation and comparison	compare_backbones(), edge_fraction()
Visualization	Comparison plotting	compare_graphs(), graph_difference()
Describe	Methods text and citations	describe_backbone()
Convert	Array and data-frame input	to_graph()
Temporal	Temporal networks: snapshots, events, and spans	convert_temporal(), flatten_temporal(), from_event_table()

Getting Started

• Installation
  • Quick install
  • Upgrading from networkx-backbone
  • Full installation
  • Hypergraph interoperability
  • Conda installation
  • Development installation
  • Requirements
• Quick Start
  • Import
  • Step 1: Create or load a weighted graph
  • Step 2: Apply a backbone method
  • Step 3: Filter to extract the backbone
  • Step 4: Evaluate the backbone
  • Disparity filter visualization
  • Other approaches
  • Next steps
• Concepts
  • What is backbone extraction?
  • Taxonomy of methods
  • The score-then-filter pattern
  • Evaluation

User Guide

• User Guide
  • Workflow
  • Choosing a Method
  • Directed Graphs
  • Weights, Units, and Transforms
  • Temporal networks: formats and conversions
  • Evaluating Backbones
  • Complexity Summary
  • Backends

Tutorials

• Tutorials
  • Start here
  • By network type
  • Going further

API Reference

• API Reference

Development

• Changelog
• Roadmap
• Contributing