A probabilistic global optimization method that explores a solution space via random perturbations and accepts changes according to a temperature-dependent probability (Metropolis criterion) so that worse moves can be accepted early to escape local optima and the temperature schedule reduces acceptance over time to concentrate near minima.

Principle

Mimic physical annealing: control exploration by a temperature parameter that gradually decreases; accept uphill moves with probability exp(-ΔE/T) to allow traversal of energy barriers, with convergence to a global optimum under an appropriately slow cooling schedule (theoretically).

Demonstration

Optimize an instance of the traveling salesman problem by randomly swapping two cities (neighbor move). At high temperature many swaps are accepted, letting the search escape local cycles; as temperature decreases fewer worse swaps are accepted, refining toward a low-cost tour. In continuous optimization, perturb vector parameters and accept according to the Metropolis rule.

Misapplication

Using too-fast cooling (temperature drops too quickly) causes premature convergence to poor local minima; using too-high initial temperature or too-long runs wastes time; choosing poor perturbation/neighborhood operators makes exploration ineffective.

Consequence

A flexible heuristic able to escape many local minima and often find high-quality approximate solutions across discrete and continuous domains; the trade-off is stochastic outcomes and tuning sensitivity to schedule and neighborhood definition.

Reversal

Greedy hill-climbing that only accepts improvements and therefore easily gets stuck in local optima, lacking the probabilistic mechanism to escape.

Boundary

Applicable to both discrete and continuous optimization problems when a neighborhood structure and an objective are defined; not guaranteed to find the global optimum under practical cooling schedules, and efficiency depends on neighborhood design and temperature schedule.

Semantic Tension

Tension between exploration (high temperature, large moves) and exploitation (low temperature, fine moves), and between simulated annealing and population-based stochastic methods (e.g., genetic algorithms) that use different mechanisms for diversity and escape.

Synthesis

Simulated annealing balances random exploration with temperature-controlled acceptance to traverse energy barriers and concentrate on low-cost regions; with careful schedule and neighborhood design it is a robust heuristic for difficult global optimization problems, though results are stochastic and sensitive to tuning.