Abstract:Optimization algorithms are very different from human optimizers. A human being would gain more experiences through problem-solving, which helps her/him in solving a new unseen problem. Yet an optimization algorithm never gains any experiences by solving more problems. In recent years, efforts have been made towards endowing optimization algorithms with some abilities of experience learning, which is regarded as experience-based optimization. In this paper, we argue that hard optimization problems could be tackled efficiently by making better use of experiences gained in related problems. We demonstrate our ideas in the context of expensive optimization, where we aim to find a near-optimal solution to an expensive optimization problem with as few fitness evaluations as possible. To achieve this, we propose an experience-based surrogate-assisted evolutionary algorithm (SAEA) framework to enhance the optimization efficiency of expensive problems, where experiences are gained across related expensive tasks via a novel meta-learning method. These experiences serve as the task-independent parameters of a deep kernel learning surrogate, then the solutions sampled from the target task are used to adapt task-specific parameters for the surrogate. With the help of experience learning, competitive regression-based surrogates can be initialized using only 1$d$ solutions from the target task ($d$ is the dimension of the decision space). Our experimental results on expensive multi-objective and constrained optimization problems demonstrate that experiences gained from related tasks are beneficial for the saving of evaluation budgets on the target problem.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to make more effective use of the experience learned from related tasks in expensive optimization problems to reduce the number of objective function evaluations, thereby improving the optimization efficiency. Specifically, the paper proposes an experience - based learning framework. By learning general experience in multiple related expensive tasks, and then applying this experience to new optimization tasks, the number of samples required for adaptive parameter adjustment in the new task is reduced, thereby reducing the optimization cost. This method is particularly suitable for optimization problems where each evaluation is very expensive, such as applications in fields like gasoline engine calibration. ### Main contributions of the paper 1. **Proposing a novel meta - learning method**: This method can learn experience from multiple related expensive tasks and generate a regression - based surrogate model, which can adapt to the adaptive parameters of the target task to approximate the fitness landscape of the target task. This surrogate model is constructed based on the deep kernel learning framework, where the Gaussian process uses the deep kernel as its covariance function. 2. **Designing an experience - based surrogate - assisted evolutionary algorithm (SAEA) framework**: This framework combines the regression - based SAEA with the above - mentioned meta - learning method. The framework contains an update strategy for adapting the surrogate model in each generation. It is worth noting that this SAEA framework is a general framework and is not designed for any specific optimization scenario, but this paper mainly focuses on multi - objective optimization and constrained optimization scenarios. 3. **Providing experimental evidence**: Demonstrating the effectiveness of the proposed experience - based SAEA framework in expensive multi - objective optimization and expensive single - objective constrained optimization. In addition, through ablation studies, some factors affecting the performance of experience - based optimization have been discovered. ### Method overview - **Meta - learning process**: Learning task - independent parameters from small data sets of multiple related tasks, which represent the domain - specific characteristics of related tasks. - **Adaptation process**: Using the experience learned from the meta - learning process to initialize the surrogate model, and then fine - tuning according to the specific data set of the target task to adapt to the target task. - **Surrogate update strategy**: Designing an update strategy to decide whether to update the surrogate model in the current iteration, ensuring the optimal update frequency of the surrogate model. Through these methods, the paper provides an effective solution that can significantly reduce the number of objective function evaluations when dealing with expensive optimization problems, thereby improving the optimization efficiency.

Experience-Based Evolutionary Algorithms for Expensive Optimization

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

Evolutionary Optimization with Simplified Helper Task for High-dimensional Expensive Multiobjective ProblemsJust Accepted

Surrogate-assisted Evolutionary Algorithms for Expensive Combinatorial Optimization: a Survey

Un-evaluated Solutions May Be Valuable in Expensive Optimization

A Hierarchical Surrogate Assisted Optimization Algorithm Using Teaching-Learning-based Optimization and Differential Evolution for High-Dimensional Expensive Problems

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

A Multiple Surrogate Assisted Decomposition-Based Evolutionary Algorithm for Expensive Multi/Many-Objective Optimization

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

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

A Novel Evolutionary Sampling Assisted Optimization Method for High-Dimensional Expensive Problems

A Surrogate-Assisted Clustering-Based Evolutionary Algorithm for Expensive Optimization.

Evolutionary Approaches to Expensive Optimisation

Evolutionary Sampling Agent for Expensive Problems

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

A Surrogate-Assisted Evolutionary Algorithm for Expensive Dynamic Multimodal Optimzation

Expensive Optimization Via Surrogate-Assisted and Model-Free Evolutionary Optimization.

A Hybrid Surrogate-Assisted Evolutionary Algorithm for Computationally Expensive Many-Objective Optimization.

A Data-Driven Evolutionary Algorithm with Multi-Evolutionary Sampling Strategy for Expensive Optimization

Accelerating Surrogate Assisted Evolutionary Algorithms for Expensive Multi-Objective Optimization Via Explainable Machine Learning

An Ensemble Surrogate-based Framework for Expensive Multiobjective Evolutionary Optimization