Working Group 2

Mixed Continuous and Discrete Optimisation

Leader: José Madeira, PT
Vice-Leader: Tobias Glasmachers, DE

WG2 brings together researchers from discrete and continuous optimisation to advance theory and algorithm design for problems where candidate solutions consist of continuous and discrete, or even combinatorial, components.

Tasks

• Developing a common analysis framework built on two unifying theories complementing each other: information geometric optimisation and drift theory.
• Advancing the design of mixed algorithms beyond ad-hoc algorithm combinations, which suffer from unclear goals and no interaction between variables. The envisioned principled designs shall implement the state of the art for each variable type, combined with the ability to model and learn dependencies between variables online. Inspiration will be drawn from Bayesian optimisation, Operations Research, and from existing discrete and continuous Randomised Optimisation Algorithms.
• Developing principled benchmark instances for the experimental evaluation and tuning of mixed black-box solvers in articulation of WG6.

Latest news