Artelys Knitro, winner of the 2019 BBComp edition!

Artelys Knitro won the expansive single objective track of the 2019 BBComp edition. This accomplishment highlights Artelys Knitro excellent performance and robustness in the field of Derivative-Free Optimization (DFO).

BBComp is a continuous black-box optimization competition where test problems are truly black boxes for participants as no assumptions can be made about the analytic form of the function to be optimized (e.g., being linear or quadratic).

The purpose of the BBComp challenge is to find the optimal objective value to a given black-box problem with a limited function evaluation budget. Artelys Knitro embeds several black-box specializations to efficiently solve such problems. The Sequential Quadratic Programming (SQP) algorithm combined with a dedicated termination criteria makes the best use of the function evaluation budget.

Such black-box formulations are frequent in several application fields, for example when optimizing networks or in blending problems. In particular, relying on external simulators usually prevents the solver from exploiting the optimization problem’s analytical formulation.