Recent Advances in Meta-features Used for Representing Black-box Single-objective Continuous Optimization
Description
This tutorial reviews and presents key advancements in representation learning of meta-features for representing optimization problem instances, algorithm instances, and their interactions in single-objective continuous black-box optimization. These representations are used for learning tasks such as algorithm selection, problem classification, and assessing complementarity between benchmark problem suites. Learning meta-features is highly relevant and important to attendees of the GECCO 2025 conference due to its potential to significantly enhance the performance and efficiency of automated optimization techniques (AutoOPT). The tutorial will explore techniques for learning meta-features specifically designed for single-objective black-box optimization, highlighting their potential applications and the possibilities of transferring them to other areas such as reinforcement learning, multi-objective optimization, etc.
Organizers
Gjorgjina Cenikj is a young researcher at the Computer Systems Department at the Jožef Stefan Institute. She is currently pursuing a PhD degree in the area of automated machine learning and representation learning for single-objective continuous optimization.
Her master thesis targeted the development of Information Extraction methods for the domain of food and nutrition.
Her main research interests include machine learning, optimization, automated machine learning, benchmarking, representation learning, natural language processing, and recommender systems.
Ana Nikolikj is a young researcher at the Computer Systems Department at the Jožef Stefan Institute in Ljubljana, Slovenia. She is working towards her PhD at the Jožef Stefan Postgraduate School, focusing on inventing methodologies to understand the behavior of single-objective numerical optimization algorithms via meta-learning. This is aimed at enhancing the process of algorithm performance prediction and algorithm selection. Her areas of interest encompass machine learning, representation learning, and methods for explainability. During her master thesis, she explored algorithm features based on explainable performance prediction models.