In either case the problem should be transformed into an unconstrained one by constructing an augmented objective function incorporating any violated constraints as penalty functions:

where ** w** is a
vector of nonnegative weighting coefficients and the vector
quantifies the magnitudes of any constraint violations.
The inverse
dependence of the penalty on temperature biases the search
increasingly heavily towards feasible space as it progresses.
Consult section for more details on the penalty function method.