Hybrid Derivative-Free Optimization for Mixed-Integer Functions
1 : LIPN - Université Paris Nord
université Paris 13
2 : IBM Watson Research Center
In this project we deviced a derivative-free method for the solution of mixed-integer functions which display combinatorial properties. We combine the principles of surrogate approximation with the difference of convex algorithm, aiming to re-use some of the features of these functions, as well to device a convergent methodology.