Simulated Annealing
A probabilistic technique for approximating the global optimum of a given function. It is often used when the search space is discrete (e.g., all tours that visit a given set of cities).
Origin
Scott Kirkpatrick, C. Daniel Gelatt Jr., and Mario Vecchi at IBM's Thomas J. Watson Research Center published "Optimization by Simulated Annealing" in Science in 1983. They drew an analogy between the physical process of annealing — slowly cooling metal to reach a low-energy state — and finding global optima in complex problems like circuit design. The method adapted the 1953 Metropolis algorithm, originally developed at Los Alamos for simulating thermodynamic systems.
Everyday Use
Imagine trying to find the lowest valley in a mountain range while blindfolded. If you only walk downhill, you might get stuck in a small dip rather than finding the deepest valley. Simulated annealing solves this by occasionally allowing uphill moves — especially early on — so you can explore more terrain before settling into the best spot. The "annealing" comes from metalworking, where slowly cooling metal produces stronger, more ordered crystals.