Transport-Free Lower Bounds for Ollivier--Ricci Curvature in Higher-Order Networks: Scope, Tightness, and Computational Applications
Organizations: Department of Computer Science and Engineering Pennsylvania State University
Abstract
We study how the Bauer--Jost--Liu curvature bound applies to hypergraph Ollivier--Ricci curvature (ORC) and how its accuracy depends on local mass and geometry. An explicit weighted-graph representation shows how the graph theorem applies. A transport construction transfers the bounds to monotone pairwise hyperedge aggregations. We analyze endpoint cancellation under laziness, quantify the exact improvement over the earlier mixture bound, and characterize the extremal geometry at fixed endpoint and overlap statistics. Further results quantify worst-case slack, stability, and exactness on isolated hyperedges. A capacity-constrained short-detour refinement uses additional geometric information and improves monotonically as the query budget increases. As a computational application, a threshold cascade preserves exact transport decisions. Numerical verification and public-hypergraph experiments examine tightness, walk dependence, and aggregation, including evaluations on 42,940 finite pairwise comparisons and 64,620 public-data instances. Together, the results specify the scope, equality mechanisms, and computational uses of BJL-type hypergraph bounds.
Figures & tables
| Method | ORC guarantee | Measures / locality | Local work | Limitation for this task |
|---|---|---|---|---|
| Transport LP | Exact optimum | General finite measures | variables; solver-dependent | Repeated optimization and distance tables |
| Entropic OT [ 23 ] | Needs feasible-plan cost or error control | General finite measures | Dense | Regularization and marginal residuals need control |
| Jost–Liu / weighted bounds [ 15 , 13 ] | One-sided in stated settings | Weighted graph walks, including loops | Local statistics | Applicability follows each theorem’s assumptions |
| Earlier [ 12 ] | One-sided | Arbitrary unit-neighborhood bases; common laziness | with adjacency queries | Lazy cancellation and short detours can cause slack |
| RS-LB [ 16 ] | One-sided | Weighted lazy -step graph walks | Shell construction and residual allocation | Requires more distance information than three statistics |
| BJL cancellation | One-sided; | Same bases as | ; after statistics | No new detour information; equals at |
| Dataset | Instances | RS-LB slack | Strict gain (%) | |||
|---|---|---|---|---|---|---|
| SHA | 44,667 | 0.199 / 0.400 | 0.112 / 0.363 | 0.085 / 0.332 | 0.000 / 0.067 | 23.8 |
| MUS | 15,345 | 0.310 / 0.531 | 0.253 / 0.525 | 0.183 / 0.454 | 0.022 / 0.096 | 49.9 |
| NDC-PC | 4,608 | 0.395 / 0.719 | 0.281 / 0.708 | 0.229 / 0.672 | 0.010 / 0.074 | 40.5 |
| Median time (s) | Solve avoidance (%) | ||||
|---|---|---|---|---|---|
| Workload | Exact OT | RS-LB | RS-LB | ||
| SHA | 0.617 | 0.288 | 0.157 | 72.7 | 90.3 |
| MUS | 1.153 | 0.973 | 0.905 | 96.4 | 100.0 |
| NDC-PC | 7.738 | 6.871 | 4.167 | 29.5 | 63.2 |