Reproducibility Data and Code for "Decoupling Boundary and Quadratic Growth: An Incomplete-Gamma Kernel for Convex Quadratic Semidefinite Optimization"
Description
This record provides the complete reproducibility material for the
manuscript “Decoupling Boundary and Quadratic Growth: An
Incomplete-Gamma Kernel for Convex Quadratic Semidefinite
Optimization.”
The archive contains the complete seed-level results and Python
implementations for 210 confirmatory runs: 120 exactly reduced spectral
experiments, 45 parameter-sensitivity experiments, and 45 dense
noncommuting nearest-correlation holdout experiments. It also includes
the generated summary tables, dependency information, fixed random
seeds, and an automated validation script.
The validation script checks the number of runs, complementarity gaps,
primal and dual residuals, positive definiteness of the dense iterates,
relative solution errors, and agreement with an independent
spectral-projection reference solver.