Exploring Nonlinear Kernels for Lipschitz Constant Estimation in Lower Bound Construction for Global Optimization

Bounds play a crucial role in guiding optimization algorithms, improving their speed and quality and providing optimality gaps. While Lipschitz constant-based lower bound construction is an effective technique, the quality of the linear bounds depends on the function’s topological properties. In this research, we improve upon this by incorporating nonlinear kernels and surrogate approximations to … Read more