Abstract:Expensive optimization problems (EOPs) are prevalent in real-world applications, where the evaluation of a single solution requires a significant amount of resources. In our study of surrogate-assisted evolutionary algorithms (SAEAs) in EOPs, we discovered an intriguing phenomenon. Because only a limited number of solutions are evaluated in each iteration, relying solely on these evaluated solutions for evolution can lead to reduced disparity in successive populations. This, in turn, hampers the reproduction operators' ability to generate superior solutions, thereby reducing the algorithm's convergence speed. To address this issue, we propose a strategic approach that incorporates high-quality, un-evaluated solutions predicted by surrogate models during the selection phase. This approach aims to improve the distribution of evaluated solutions, thereby generating a superior next generation of solutions. This work details specific implementations of this concept across various reproduction operators and validates its effectiveness using multiple surrogate models. Experimental results demonstrate that the proposed strategy significantly enhances the performance of surrogate-assisted evolutionary algorithms. Compared to mainstream SAEAs and Bayesian optimization algorithms, our approach incorporating the un-evaluated solution strategy shows a marked improvement.

What problem does this paper attempt to address?

This paper attempts to solve the problem of inefficiency of traditional evolutionary algorithms (EAs) in expensive optimization problems (EOPs) due to the large amount of resources required for each evaluation of a single solution. Specifically: 1. **Limitations of traditional methods**: - In expensive optimization problems, a large amount of computational resources or time is required for each evaluation of a solution. - Traditional evolutionary algorithms rely on the real - evaluation of all solutions, which is impractical in expensive optimization problems because the evaluation cost is too high. - In order to reduce the number of evaluations, surrogate - assisted evolutionary algorithms (SAEAs) use surrogate models to predict the quality of solutions, thereby reducing the need for real - evaluation. 2. **Problems discovered**: - In SAEAs, since only a limited number of solutions can be evaluated in each iteration, relying solely on these evaluated solutions for evolution will lead to a decrease in population diversity, which in turn affects the ability of the reproduction operator to generate better solutions and ultimately slows down the convergence speed of the algorithm. - This phenomenon is known as "standing still" in the literature, that is, the algorithm cannot significantly improve the distribution of new solutions during the evolution process, resulting in a slow optimization process. 3. **Proposed solutions**: - The paper proposes a new strategy, which combines high - quality un - evaluated solutions predicted by the surrogate model and evaluated solutions in the selection stage to generate the next - generation population. - Specifically, although un - evaluated solutions have not been directly evaluated in reality, they still have potential value, which can increase the diversity of the population, improve the effectiveness of the reproduction operator, and thus accelerate the optimization process. 4. **Specific implementation**: - This strategy has been specifically implemented on three typical reproduction operators: genetic algorithm (GA), differential evolution (DE), and estimation of distribution algorithm (EDA). - The experimental results show that this strategy significantly improves the performance of surrogate - assisted evolutionary algorithms, outperforming the mainstream SAEAs and Bayesian optimization algorithms. In summary, this paper aims to solve the problem of inefficiency of traditional evolutionary algorithms in expensive optimization problems by introducing high - quality un - evaluated solutions, and verifies the effectiveness of this strategy through experiments.

Un-evaluated Solutions May Be Valuable in Expensive Optimization

Enhancing SAEAs with Unevaluated Solutions: A Case Study of Relation Model for Expensive Optimization

A review of surrogate-assisted evolutionary algorithms for expensive optimization problems

A composite surrogate-assisted evolutionary algorithm for expensive many-objective optimization

Experience-Based Evolutionary Algorithms for Expensive Optimization

A surrogate-ensemble assisted expensive many-objective optimization

An Ensemble Surrogate-based Framework for Expensive Multiobjective Evolutionary Optimization

A Surrogate-Assisted Evolutionary Algorithm for Seeking Multiple Solutions of Expensive Multimodal Optimization Problems

Multiple surrogates and offspring-assisted differential evolution for high-dimensional expensive problems

Efficient Surrogate Modeling Method for Evolutionary Algorithm to Solve Bilevel Optimization Problems

An activity level based surrogate-assisted evolutionary algorithm for many-objective optimization

Surrogate-assisted Autoencoder-embedded Evolutionary Optimization Algorithm to Solve High-dimensional Expensive Problems

A Surrogate-Assisted Evolutionary Algorithm for Expensive Dynamic Multimodal Optimzation

A bagging-based surrogate-assisted evolutionary algorithm for expensive multi-objective optimization

Surrogate Assisted Evolutionary Algorithm for Medium Scale Expensive Multi-Objective Optimisation Problems

Surrogate-assisted evolutionary algorithm for expensive constrained multi-objective discrete optimization problems

A Supervised Surrogate-Assisted Evolutionary Algorithm for Complex Optimization Problems

Computationally Expensive High-Dimensional Multiobjective Optimization via Surrogate-Assisted Reformulation and Decomposition

A Hierarchical and Ensemble Surrogate-Assisted Evolutionary Algorithm With Model Reduction for Expensive Many-Objective Optimization

Efficient hierarchical surrogate-assisted differential evolution for high-dimensional expensive optimization