Category: Physics

Summary: Testing whether a moderate delay in the annealing schedule produces a better optimization window than either immediate or very late annealing in a spin-glass landscape.

Annealing schedules in frustrated systems are often tuned by trial and error, even though too-rapid and too-delayed cooling can fail for different reasons. This experiment asks whether there is an interior delay window that yields better outcomes than either extreme in a spin-glass-style search problem.

The script compares several delay values, computes the gain achieved by each one, and summarizes whether the best delay lies in the middle of the tested range. It also records acceptance rates and final magnetization to distinguish improved exploration from simple freezing.

That makes the project a window-finding problem rather than a single-parameter benchmark. The main target is whether delayed annealing produces a genuine interior optimum, which would point to timing-sensitive search dynamics in rough energy landscapes.

Method: Repeated spin-glass annealing runs comparing discrete delay values and summarizing interior-window gains, acceptance, and magnetization.

What is measured: Best delay, gain at each tested delay, window index, acceptance rate, absolute magnetization, and the fraction of trials with an interior optimum.