AutoDesign4EC: Automated Algorithm Design for
Evolutionary Computation

Scope and aim

Automating the design of new heuristics that are suitable for a given optimization problem instance is a core research objective in computational intelligence. It is a demanding task, as it requires a solid understanding of the complex interactions between the algorithm configuration space and the problem space. Many different techniques for the automated selection, configuration, and design of black-box optimization algorithms co-exist, in a yet very unorganized way across different sub-domains of evolutionary computation, machine learning, and operations research.

This special session intends to bring together researchers interested in the automated design of optimization algorithms for black-box optimization. No restriction will be made on the particular application domain, i.e, we are equally interested in combinatorial/discrete/numerical/mixed-integer optimization, noisy/non-noisy settings, static/dynamic problems, single-/multi-objective optimization, different performance metrics, etc. The unifying theme are the approaches that are used to automate algorithms’ design.

Topics of Interest

• Supervised ML for automated algorithm selection, configuration, and design
• Unsupervised ML for automated algorithm selection, configuration, and design
• Explainable machine learning for automated algorithm selection, configuration, and design
• Meta-learning
• Transfer of approaches between machine learning and optimization
• Taxonomies/ontologies for describing the algorithm instance space
• Complementary analysis of different benchmarking datasets
• Representation learning techniques for structured, unstructured, and graph data
• Exploratory landscape analysis for characterizing problem or algorithm space
• Feature selection, ranking and sensitivity analysis
• Representation learning applied on landscape data
• Representation learning applied on performance data
• Improving understanding of data (landscape and/or performance) through visualization techniques