Abstract:The selection of the most appropriate algorithm to solve a given problem instance, known as algorithm selection, is driven by the potential to capitalize on the complementary performance of different algorithms across sets of problem instances. However, determining the optimal algorithm for an unseen problem instance has been shown to be a challenging task, which has garnered significant attention from researchers in recent years. In this survey, we conduct an overview of the key contributions to algorithm selection in the field of single-objective continuous black-box optimization. We present ongoing work in representation learning of meta-features for optimization problem instances, algorithm instances, and their interactions. We also study machine learning models for automated algorithm selection, configuration, and performance prediction. Through this analysis, we identify gaps in the state of the art, based on which we present ideas for further development of meta-feature representations.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: how to automatically select the most appropriate algorithm instance for single - objective continuous black - box optimization problems. Specifically, the paper focuses on improving the algorithm selection method through the representation learning of meta - features. ### Detailed Interpretation 1. **Problem Background**: - Different optimization problem instances and different performance metrics require different algorithm instances to reach the optimal solution. - In computationally complex problems, usually no single algorithm instance can perform best consistently. Instead, there is often a set of algorithm instances with complementary advantages. - Therefore, selecting the algorithm instance that is most suitable for a specific problem instance (i.e., the algorithm selection problem) becomes crucial. 2. **Research Motivation**: - The goal of algorithm selection is to utilize the different performance of different algorithm instances on different sets of problem instances to find the optimal solution. - However, finding the optimal algorithm for a new problem instance is a challenging task and has attracted a great deal of research interest in recent years. 3. **Main Contributions**: - The paper reviews algorithm selection in the field of single - objective continuous black - box optimization. - It focuses on introducing the latest progress in meta - feature representation learning for optimization problem instances, algorithm instances, and their interactions. - It studies machine - learning models for automated algorithm selection, configuration, and performance prediction. - It identifies the deficiencies in existing research and proposes directions for further development. ### Formula Examples In optimization problems, commonly used evaluation metrics include two scenarios: fixed - budget and fixed - target: - **Fixed - Budget**: \[ \text{Budget} = B, \quad \text{Solution Quality} = f(\mathbf{x}) \] where \( B \) is the predefined budget and \( f(\mathbf{x}) \) is the quality of the solution recommended by the algorithm. - **Fixed - Target**: \[ \text{Target Quality} = T, \quad \text{Required Budget} = B(f(\mathbf{x}) \geq T) \] where \( T \) is the desired solution quality and \( B(f(\mathbf{x}) \geq T) \) is the budget required to reach or exceed this quality. ### Summary This paper aims to explore how to more effectively select the most appropriate algorithm instance for single - objective continuous black - box optimization problems by reviewing existing meta - feature representation learning methods. Through in - depth analysis of problem instances, algorithm instances, and their interactions, the paper provides valuable insights and directions for future research.

A Survey of Meta-features Used for Automated Selection of Algorithms for Black-box Single-objective Continuous Optimization

Benchmarking Feature-based Algorithm Selection Systems for Black-box Numerical Optimization

A Survey for Study of Feature Selection Algorithms

PS-AAS: Portfolio Selection for Automated Algorithm Selection in Black-Box Optimization

Multiclass feature selection with metaheuristic optimization algorithms: a review

Automated Design of Metaheuristic Algorithms: A Survey

Spotted Hyena Optimization Algorithm with Simulated Annealing for Feature Selection

Toward Automated Algorithm Design: A Survey and Practical Guide to Meta-Black-Box-Optimization

A Survey of Feature Selection Algorithm

A meta-feature selection method based on the Auto-sklearn framework

Impact of Training Instance Selection on Automated Algorithm Selection Models for Numerical Black-box Optimization

A Comprehensive Survey on Feature Selection with Grasshopper Optimization Algorithm

Automated Selection of Evolutionary Multi-objective Optimization Algorithms.

Benchmarking for Metaheuristic Black-Box Optimization: Perspectives and Open Challenges

A Brief Survey on Nature-Inspired Metaheuristics for Feature Selection in Classification in This Decade.

A survey on new generation metaheuristic algorithms

An Overview and Comparison of Selected State-of-the-Art Algorithms Inspired by Nature

Sequential vs. Integrated Algorithm Selection and Configuration: A Case Study for the Modular CMA-ES

A Literature Survey and Empirical Study of Meta-Learning for Classifier Selection

Comparative Analysis of Selection Hyper-Heuristics for Real-World Multi-Objective Optimization Problems

Per-run Algorithm Selection with Warm-starting using Trajectory-based Features