Black-Box Optimization with Hexaly, by Emeline Tenaud

Event Simulation Optimization

Louisa Senior Technical Writer November 4, 2024

Emeline Tenaud, Senior Optimization Scientist at Hexaly, discussed Black-Box Optimization with Hexaly at the 44th Parisian Day of Operations Research (JFRO 44) on November 26, 2024, at Sorbonne Université in Paris. Here is the summary of her talk:

In black-box optimization problems, the analytical form of the objective function and constraints is unknown, and evaluating new solutions can be costly in terms of time and memory. Hexaly provides a surrogate modeling method for optimizing these problems using radial basis function (RBF) models, which helps to reduce the number of evaluations by carefully selecting the points to evaluate. The algorithm combines exploration and intensification phases to make the most of a limited evaluation budget. Additionally, Hexaly handles both analytical and black-box constraints that are often found in industrial applications and can optimize multiple black-box objectives simultaneously. This presentation presents the approach implemented in Hexaly and illustrates its effectiveness in real-world applications.

Black-Box Optimization enables Hexaly’s users to couple Mathematical Optimization with Numerical or Discrete-Event Simulation and Mathematical Optimization with Machine Learning in a straightforward way. Discover more about Hexaly’s Black-Box Optimization capabilities in Hexaly’s Doc Center.