simulated annealing

Gold definitionUpdated Apr 2, 2026

Definition

Simulated annealing is a metaheuristic optimization algorithm inspired by the annealing process in metallurgy. It explores a solution space to find a good approximation to a global optimum in a large search space, especially for combinatorial optimization problems.

At a glance

Executive summary

Simulated annealing is an optimization method that finds good solutions to hard problems by mimicking how metals cool and crystallize. It starts by broadly searching for solutions and gradually narrows down to the best one, even accepting slightly worse options occasionally to avoid getting stuck. This makes it great for complex problems where finding the absolute perfect answer is too slow.

TL;DR

It's an algorithm that finds good solutions to difficult problems by randomly trying options and slowly refining them, like cooling a hot metal.

Key points

Iteratively explores solutions, accepting better ones and occasionally worse ones with decreasing probability (temperature).
Finds approximate global optima for complex combinatorial optimization problems where exact solutions are intractable.
Used by researchers in operations research, computer science, physics, engineering, and increasingly in probabilistic computing and machine learning.
Unlike greedy algorithms that always pick the best immediate option and can get stuck in local optima, SA can escape local optima by accepting worse solutions probabilistically.
Integration with emerging hardware like p-bits for probabilistic computing, leveraging device variability for enhanced performance, and GPU acceleration for large-scale problems.

Use cases

VLSI Design: Optimizing the placement of components on integrated circuits to minimize wire length and improve performance.
Traveling Salesperson Problem (TSP): Finding the shortest possible route that visits each city exactly once and returns to the origin city.
Job Shop Scheduling: Optimizing the sequence of operations on machines to minimize total completion time or maximize throughput.
Protein Folding: Predicting the 3D structure of a protein from its amino acid sequence by minimizing its energy configuration.
Image Segmentation: Grouping pixels into regions to simplify and/or change the representation of an image into something more meaningful and easier to analyze.

Also known as

SA, Fast Simulated Annealing, Adaptive Simulated Annealing, Parallel Simulated Annealing