Maximum entropy modeling of Optimal Transport: the sub-optimality regime and the transition from dense to sparse networks

Abstract

We present a bipartite network model that captures intermediate stages of optimization by blending the Maximum Entropy approach with Optimal Transport. In this framework, the network's constraints define the total mass each node can supply or receive, while an external cost field favors a minimal set of links, driving the system toward a sparse, tree-like structure. By tuning the control parameter, one transitions from uniformly distributed weights to an optimal transport regime in which weights condense onto cost-favorable edges. We quantify this dense-to-sparse transition, showing with numerical analyses that the process does not hinge on specific assumptions about the node-strength or cost distributions. Finite-size analysis confirms that the results persist in the thermodynamic limit. Because the model offers explicit control over the degree of sub-optimality, this approach lends to practical applications in link prediction, network reconstruction, and statistical validation, particularly in systems where partial optimization coexists with other noise-like factors.

arXiv comment30 pages, 7 figures, submitted to Communications Physics

References in corpus (11)

• The statistical mechanics of networks
• The Statistical Physics of Real-World Networks
• Damage and fluctuations induce loops in optimal transport networks
• The entropy of randomized network ensembles
• Fluctuations and redundancy in optimal transport networks
• Maximum likelihood: extracting unbiased information from complex networks
• Network Structures from Selection Principles
• Fast and scalable likelihood maximization for Exponential Random Graph Models with local constraints
• Reconstructing networks
• The Grand Canonical ensemble of weighted networks
• Inferring comparative advantage via entropy maximization