This repository contains the Python source code and the stochastic simulation framework accompanying the manuscript: "Scale-invariant macroscopic observables in Pareto-distributed networks via thermodynamic aggregation". The archive provides the complete mathematical and computational architecture required to reproduce the topological coarse-graining experiments on heavy-tailed environments. The numerical engine computes the exact 1D Wasserstein metric to evaluate the topological information loss incurred by linear operators (arithmetic mean) versus non-linear, finite-temperature thermodynamic aggregation mechanisms. The repository includes the exact numerical implementations used to generate:- The macroscopic dilution effect and local partition function phase transitions.- The probability density functions (PDF) mapping macroscopic noise.- The topological resilience envelopes across the structural sparsity spectrum (from Gaussian limits to Zipfian criticality).- The finite-size scaling phase diagrams (up to N=2,000). All routines are fully vectorized and implement numerically stable log-space generation for extreme Pareto topologies to prevent floating-point overflow at extreme sparsity regimes (alpha < 0.05).
Corentin Guigot (Thu,) studied this question.