Simulated Annealing
Simulated annealing is a paradigm method to solve optimization problems of many types. It is a meta-heuristic model which allows worse moves so as to escape from the local optima. It is a memory-less algorithm: the algorithm does not use any information gathered during the search. It is applied for both combinatorial and continuous optimization problems. It is an iterative process, which is both simple and easy to implement.
One of the key features of this process is that by providing a platform for hill climbing moves, a means to escape the local optima is created (hill climbing moves are those moves that worsen the objective function value)-It does this with the hope of finding a global optimum.
This optimization technique is based on the Metropolis acceptance criterion, which shadows the way a thermodynamic system moves from a current state to a candidate solution where energy content is minimized. Based on the acceptance probability, the candidate solution is accepted or rejected. Simulated annealing is frequently used when there is a discrete search space (the number of tours that visit a set of cities).This technique is sometimes more effective than exhaustive enumeration, when the objective is to find a good solution which is acceptable, in a fixed span of time (as compared to the solution)
The motivation comes from annealing in metallurgy.
Physical Annealing
Annealing refers to the process by which a material is alternatively heated and cools, in order to reduce the defects in the crystals by increasing the size. Heating and cooling directly affects the thermodynamic free energy and the temperature of the material.
Simulated annealing and physical annealing
The idea of gradual cooling is applied in the Simulated Annealing algorithm as a gradual reduction in the chance of accepting a worse solution as it searches the space