General properties of simulated annealing

In addition to computational efficiency analyzed above the following general properties merit mention, a. Simplicity. 

The basic simulated annealing algorithm does not require much more than a dozen lines of code (not including evaluation of the objective function) and the optimization constraints are easily handled by rejecting candidate points that violate them. 

Prescription of different fimctional dependencies of p, incorporation of different stretchings for Cerent variables, various recipes for pseudorandom step selection, and specifications of various termination criteria ensure that the basic simulated annealing algorithm can be modified to meet special requirements of most applications. 

The method is applicable to nonconvex objective functions with multiple optima and to nondifferentiable (discontinuous) functions and it may be used for discrete (combinatorial) and continuous variables optimization. 

We have used simulated annealing to determine optimal designs of biological experiments [4] and optical laises [9,10], optimal deployments of missile interceptors [11], optimal allocations of resources in orbital engagements [12] and optimal molecular conformations [13]. 

---