PlatMetaX: An Integrated MATLAB platform for Meta-Black-Box Optimization

TL;DR

PlatMetaX proposes a unified MATLAB platform for Meta-Black-Box Optimization by integrating RL-assisted meta-learning with a broad suite of base-optimizers and problem sets drawn from PlatEMO. It introduces modular templates (Environment, Base-Optimizer, Meta-Optimizer) and a standardized training/testing workflow to develop adaptive algorithms across single- and multi-objective problems, leveraging benchmarks like BBOB2009. The study demonstrates meta-optimizers such as DDPG_DE_F and DQN_DE_MS, highlighting that dynamic mutation-strategy selection (DQN_DE_MS) often improves performance and transferability, while RL-based scaling of $F$ (DDPG_DE_F) can underperform on unseen problems. Key contributions include transferability and generalization metrics and a graphical workflow compatible with PlatEMO's GUI, enabling rapid experimentation and comparisons. The platform aims to accelerate automated algorithm design in optimization with modularity, extensibility, and potential future enhancements like LLM integration and richer logging.

Abstract

The landscape of optimization problems has become increasingly complex, necessitating the development of advanced optimization techniques. Meta-Black-Box Optimization (MetaBBO), which involves refining the optimization algorithms themselves via meta-learning, has emerged as a promising approach. Recognizing the limitations in existing platforms, we presents PlatMetaX, a novel MATLAB platform for MetaBBO with reinforcement learning. PlatMetaX integrates the strengths of MetaBox and PlatEMO, offering a comprehensive framework for developing, evaluating, and comparing optimization algorithms. The platform is designed to handle a wide range of optimization problems, from single-objective to multi-objective, and is equipped with a rich set of baseline algorithms and evaluation metrics. We demonstrate the utility of PlatMetaX through extensive experiments and provide insights into its design and implementation. PlatMetaX is available at: \href{https://github.com/Yxxx616/PlatMetaX}{https://github.com/Yxxx616/PlatMetaX}.

Figures (10)

• Figure 1: Bi-level optimization framework of MetaBBO.
• Figure 2: The architecture of PlatMetaX.
• Figure 3: The flowchart of MetaBBO.
• Figure 4: UML class of PlatMetaX. Users can define own methods from provided templates highlighted in yellow to enable polymorphism. Classes in blue box are the official classes of MATLAB, while classes in orange box are the original classes of PlatEMO.
• Figure 5: Unified working logic of three kinds of meta-optimizers, where the part in dashed box represents one step of the meta-optimizer. Different meta-optimizers can be implemented via the action generation strategies and specific step method which update the mapping policy.