This research introduces HEQP, a novel method to enhance solution quality in quadratically constrained quadratic programs, suggesting its effectiveness.
Key Points
The study aims to develop an advanced optimization framework for solving large-scale QCQPs using machine learning techniques.
Introduced a hypergraph neural network to predict optimal solutions without reliance on parametric models.
Implemented an evolutionary large neighborhood search (Evo-LNS) for solution refinement and application of crossover techniques.
Demonstrated equivalence to the interior-point method for quadratic programming.
HEQP outperformed existing solvers like Gurobi and SCIP in solution quality.
Demonstrated improved time efficiency in solving large-scale QCQPs on benchmark problems.