Handling Box, Linear and Quadratic-Convex Constraints for Boundary Optimization with Differential Evolution Algorithms

We propose and test the performance of an implicit strategy to handle box, linear and quadratic convex constraints, based on changing the search space from points to directions, suitable to be easily implemented in combination with differential evolution (DE) algorithms for the boundary optimization of a generic continuous function. In particular, we see that DE can be efficiently implemented to find solutions on the boundary of box constraints, linear inequality constraints and quadratic convex constraints, for which the feasible set is convex and bounded. The computational results are performed on different classes of minimization problems.