For example, an algorithm called “simulated annealing” starts with an almost random search, willing to try any change, good or bad. Then it slowly becomes fussier and fussier about what changes it will accept, until eventually it has turned into a rigid search for small step-by-step improvements.
