Biased Graph¶
A graph with selected balanced cycles satisfying the theta rule: no theta subgraph has exactly two balanced cycles.
Core Idea¶
A biased graph is a graph with some simple cycles designated balanced. The selection must obey the theta rule: among the three cycles formed by three internally disjoint paths between the same two vertices, exactly two cannot be balanced. If two are balanced, the third must be. This is a formal graph structure, not a claim about social or statistical bias.
Scope of Application¶
Signed graphs generate a balanced-cycle class using even parity of negative edges. Group-gain graphs generate one when the product of oriented edge gains around a cycle is the group identity. Both satisfy the theta rule. Biased graphs let graph and matroid theorists reason about the balance pattern without requiring one particular label system. Frame and lift matroids can be derived from the result but are not required input.
Clarity¶
A graph with arbitrarily marked cycles is not necessarily biased: exactly two marked cycles in one theta invalidate it. Signs or gains are optional ways to produce the class, not part of the general definition. A signed graph is one specialization, and a matroid derived from balanced cycles is a different object.
Manages Complexity¶
The selected cycle class retains the relevant balance pattern while forgetting many details of its edge labels. A result expressed in terms of theta-consistent cycles can apply across signed and gain realizations. What was forgotten cannot be recovered: the bare biased graph does not identify the original sign or gain on each edge.
Abstract Reasoning¶
Write \((G,\mathcal B)\) for a graph and its selected cycles. Inspect every theta's three path-pair cycles; reject the selection if precisely two lie in \(\mathcal B\). Signed parity and group-gain identity products each satisfy this test because two consistent path comparisons force the third. Only after the balance class is established should one derive frame or lift matroid properties.
Knowledge Transfer¶
In another graph-theoretic setting, ask which conclusions depend only on the graph and theta-closed cycle selection and which need a chosen labeling group or a particular matroid construction. The staged strict DAG parent is live Network, which supplies node–edge structure but not the additional balanced-cycle axiom.
Relationships to Other Abstractions¶
Current abstraction Biased Graph Domain-specific
Parents (1) — more general patterns this builds on
-
Biased Graph is a kind of Network Prime
A biased graph is a vertex–edge connection structure with the additional theta-consistent balanced-cycle class.
Hierarchy path (1) — routes to 1 parentless root
- Biased Graph → Network → Reservoir-Flux Network → Conservation Laws → Invariance
Neighborhood in Abstraction Space¶
Biased Graph sits in a sparse region of the domain-specific corpus (79th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.
Family — Data Structures & Graph Variants (17 abstractions)
Nearest neighbors
- Complement graph — 0.83
- Chordal bipartite graph — 0.83
- Graph Toughness — 0.82
- Polygon — 0.82
- Twin-width — 0.82
Computed from structural-signature embeddings · 2026-10-08