Pathfinder 8: A Bayesian Network Graph with 109 Nodes

A sample from the Pathfinder Bayesian Network, a discrete, very-large network used as a machine learning benchmark. The network contains 109 nodes, 195 arcs, and 72,079 parameters, with an average Markov blanket size of 3.82. It was authored by D. Heckerman, E. Horwitz, and B. Nathwani, based on research published in 1992.

Use Cases

• Benchmarking graph classification algorithms based on the network's 109-node structure.
• Testing probabilistic inference methods based on the network's 72,079 parameters.
• Studying graph topology properties based on metrics like average degree (3.58) and maximum in-degree (5).
• Evaluating structure learning algorithms for Bayesian networks based on the known ground-truth graph.

Strengths

• Well-defined graph structure with 109 nodes and 195 arcs, providing a concrete benchmark topology.
• Extensive parameterization with 72,079 parameters, enabling complex probabilistic modeling.
• Clear academic provenance, authored by known researchers and cited in a 1992 publication.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for certain learning tasks.
• Last update date is unknown; freshness unverified.

Provenance

Source

bnlearn Bayesian Network Repository, originally from the Pathfinder Project by D. Heckerman, E. Horwitz, and B. Nathwani.

Collection Method

Likely a synthetic or expert-constructed Bayesian network for diagnostic reasoning.

Time Range

Based on research published in 1992.

Freshness

Last updated date is unknown.

Geography

null