T
Twiki
Text
Quantization-Based Global Optimization via Langevin Dynamics and the Witten-Laplacian
A stochastic and quantum-mechanical analysis of quantization-based global optimization, where iteratively quantizing an objective function shrinks the level set that contains its saddle points and local minima toward the global optimum. Treating the uniform quantization error as i.i.d. white noise yields an overdamped Langevin stochastic differential equation for the dynamics, and the associated Witten-Laplacian explains the method's ability to escape local minima, linking it to thermodynamic (simulated annealing) and quantum (quantum annealing) global optimization.