Article
Computer Science, Artificial Intelligence
Qiqi Liu, Yaochu Jin, Martin Heiderich, Tobias Rodemann
Summary: A new surrogate-assisted evolutionary algorithm is proposed in this study to handle expensive irregular multi-objective optimization problems. The algorithm balances convergence and diversity by adapting reference vectors and implementing a surrogate management strategy, effectively taking irregularity of the Pareto front into account.
KNOWLEDGE-BASED SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Xiwen Cai, Tao Zou, Liang Gao
Summary: This study proposes a surrogate-assisted multi-objective evolutionary algorithm that integrates multiple surrogate-assisted strategies to improve the optimization efficiency of computationally expensive multi-objective problems. The algorithm utilizes a surrogate-assisted penalty-based boundary intersection infill criterion and an operator-repeated offspring creation strategy for global search and diversity of Pareto optimal solutions. In addition, an improved surrogate-based multi-objective local search method is introduced to accelerate convergence speed. Experimental results demonstrate the superior performance of the proposed algorithm compared to state-of-the-art approaches.
APPLIED SOFT COMPUTING
(2023)
Article
Computer Science, Artificial Intelligence
Fan Li, Liang Gao, Akhil Garg, Weiming Shen, Shifeng Huang
Summary: This paper proposes a surrogate-assisted dominance-based multi-objective evolutionary algorithm, which can efficiently solve multi-objective computationally expensive problems with medium dimensions. By using convergence and diversity criteria collaboratively, the algorithm enhances the exploration of the population and improves the accuracy of surrogate models.
SWARM AND EVOLUTIONARY COMPUTATION
(2021)
Article
Computer Science, Artificial Intelligence
Wenxin Wang, Huachao Dong, Peng Wang, Jiangtao Shen
Summary: This paper proposes a bi-indicator-based surrogate-assisted evolutionary algorithm (BISAEA) for solving computationally expensive multi-objective optimization problems (MOPs). BISAEA utilizes a Pareto-based bi-indicator strategy and a radius-based function (RBF) model to approximate objective values. It also incorporates a one-by-one selection strategy based on angles and Pareto dominance to improve diversity. Experimental results show that BISAEA achieves high efficiency and a good balance between convergence and diversity. Application of BISAEA to a multidisciplinary optimization problem further demonstrates its superior performance on computationally expensive engineering problems.
COMPLEX & INTELLIGENT SYSTEMS
(2023)
Article
Computer Science, Information Systems
Hao Wang, Chaoli Sun, Gang Xie, Xiao-Zhi Gao, Farooq Akhtar
Summary: Surrogate-assisted multi-objective evolutionary algorithms have been focused on to solve expensive multi-objective problems. Gaussian process models are proposed for performance indicators instead of objective functions. The efficiency of the approach is validated on test suites and a real-world optimization problem, and it is found to be competitive compared to peer algorithms for expensive many-objective problems.
INFORMATION SCIENCES
(2023)
Article
Computer Science, Artificial Intelligence
Jinglu Li, Peng Wang, Huachao Dong, Jiangtao Shen, Caihua Chen
Summary: In this article, a classification surrogate-assisted multi-objective evolutionary algorithm (CSA-MOEA) is proposed for solving expensive optimization problems. The algorithm adopts a classification tree as the surrogate model and obtains valuable solutions through two infilling strategies.
KNOWLEDGE-BASED SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Shufen Qin, Chaoli Sun, Qiqi Liu, Yaochu Jin
Summary: In surrogate-assisted multi-/many-objective evolutionary optimization, an efficient model management strategy is highly challenging due to the complex tradeoff between different objectives and accumulated uncertainty in the approximation of objective functions. This article proposes building surrogate models for each objective function and using a Gaussian process model to approximate a performance indicator. The experimental results demonstrate the competitiveness of the proposed method in comparison to state-of-the-art surrogate-assisted evolutionary algorithms.
IEEE TRANSACTIONS ON EVOLUTIONARY COMPUTATION
(2023)
Article
Computer Science, Artificial Intelligence
Yi Zhao, Jian Zhao, Jianchao Zeng, Ying Tan
Summary: This paper proposes a two-stage infill strategy and surrogate-ensemble assisted optimization algorithm for solving expensive many-objective optimization problems. Experimental results demonstrate the superiority of this algorithm in solving computationally expensive many-objective optimization problems.
COMPLEX & INTELLIGENT SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Shufen Qin, Chaoli Sun, Farooq Akhtar, Gang Xie
Summary: This paper proposes a Kriging-assisted many-objective optimization algorithm that uses two infill sampling criteria to adaptively select new solutions for objective function evaluations, improving historical models. The proposed algorithm shows competitive performance on most optimization problems compared to four classical surrogate-assisted multi-objective evolutionary algorithms.
Article
Computer Science, Artificial Intelligence
Mingyuan Yu, Jing Liang, Zhou Wu, Zhile Yang
Summary: Surrogate-assisted evolutionary algorithms have gained significant attention in solving expensive optimization problems. In this study, a twofold infill criterion-driven heterogeneous ensemble surrogate-assisted neighborhood field optimization algorithm (HESNFO) is proposed, which takes into account the diversity and accuracy of surrogates to speed up the optimization process.
KNOWLEDGE-BASED SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Yi Zhao, Jianchao Zeng, Ying Tan
Summary: The proposed method combines reference vector guided evolutionary algorithm and radial basis function networks to optimize individuals and introduces an infill strategy, showing competitive performance in solving computationally expensive many-objective optimization problems.
APPLIED SOFT COMPUTING
(2021)
Article
Computer Science, Artificial Intelligence
Yi Zhao, Chaoli Sun, Jianchao Zeng, Ying Tan, Guochen Zhang
Summary: This paper proposes a method to train multiple surrogate models to assist many-objective optimization algorithm for solving expensive many-objective problems. Experimental results show that this method is competitive with other surrogate-assisted evolutionary algorithms within a limited computational budget.
KNOWLEDGE-BASED SYSTEMS
(2021)
Article
Computer Science, Artificial Intelligence
Qinghua Gu, Yufeng Zhou, Xuexian Li, Shunling Ruan
Summary: Model management is crucial for surrogate-assisted evolutionary algorithms to tackle computationally expensive optimization problems. The proposed algorithm in this paper, based on radial space division, outperforms commonly used surrogate-assisted evolutionary algorithms in benchmark and automobile structure design problems.
APPLIED SOFT COMPUTING
(2021)
Article
Computer Science, Artificial Intelligence
Dong Han, Wenli Du, Xinjie Wang, Wei Du
Summary: The study introduces a surrogate-assisted decomposition-based evolutionary algorithm that considers the balance between exploration and exploitation by incorporating distribution information of weight vectors and population, as well as develops a replacement strategy to limit model training time. Empirical results demonstrate competitive performance of the algorithm, with superior outcomes when applied to real-world optimization problems.
SWARM AND EVOLUTIONARY COMPUTATION
(2022)
Article
Computer Science, Artificial Intelligence
Qinghua Gu, Qian Wang, Neal N. Xiong, Song Jiang, Lu Chen
Summary: A surrogate-assisted evolutionary algorithm is proposed in this paper for solving expensive constrained multi-objective discrete optimization problems. By embedding random forest models and an improved stochastic ranking strategy, the algorithm makes significant progress in optimization efficiency and candidate solution quality.
COMPLEX & INTELLIGENT SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Shulei Liu, Handing Wang, Wei Peng, Wen Yao
Summary: This study proposes a surrogate-assisted evolutionary algorithm (SAEA) for expensive feature selection problems. By employing parallel random grouping and a constraint-based sampling strategy, the algorithm effectively optimizes high-dimensional discrete decision variables. Experimental results demonstrate that the proposed algorithm outperforms traditional and ensemble feature selection methods on multiple datasets.
IEEE TRANSACTIONS ON EVOLUTIONARY COMPUTATION
(2022)
Article
Computer Science, Artificial Intelligence
Shulei Liu, Handing Wang, Wen Yao
Summary: This paper proposes a novel surrogate-assisted evolutionary algorithm (SAEA) for addressing computationally expensive and noisy combinatorial multi-objective optimization problems. The algorithm uses an averaging method for denoising and constructs multi-fidelity surrogate models based on averaged evaluation results. The fidelity level of surrogate models is determined by the number of independent repeated evaluations. The hypervolume indicator is employed as a trigger to increase the fidelity level during the optimization process. Additionally, a lightweight local search method, the semi-variable neighborhood search, is proposed to enhance global search efficiency in discrete decision spaces.
APPLIED SOFT COMPUTING
(2022)
Article
Computer Science, Artificial Intelligence
Yongcun Liu, Handing Wang
Summary: This study proposes a novel algorithm that combines global and local search strategies to address the challenge of multiple disconnected regions in the search space. The algorithm achieves competitive results with only hundreds of function evaluations and can handle mixed-variable optimization problems. The global module uses hybrid evolutionary operators and a Gower distance based surrogate model, while the local module performs competitive switching in different local regions and improves evaluation accuracy with local surrogate models. The algorithm is demonstrated to be effective through artificial benchmark tests and convolutional neural network hyperparameter optimization problems.
APPLIED SOFT COMPUTING
(2023)
Article
Computer Science, Artificial Intelligence
Chao Li, Wen Yao, Handing Wang, Tingsong Jiang
Summary: It has been found that deep neural networks are vulnerable to adversarial examples for several years. Existing transfer-based methods have weak transferability for black-box models and sparse attacks mainly focus on the number of attacked pixels without restricting the size of perturbations. To address these issues, this study proposes a transfer-based sparse attack method that improves transferability through adaptive momentum variance and refining perturbation mechanism, and uses a class activation map to explore the relationship between the number of perturbed pixels and attack performance.
PATTERN RECOGNITION
(2023)
Article
Computer Science, Artificial Intelligence
Zhening Liu, Handing Wang
Summary: This paper proposes two strategies (training data augmentation and sampling pool expansion) to be incorporated in the Kriging-assisted reference vector guided evolutionary algorithm, aiming to solve expensive dynamic multi-objective optimization problems. The experiment results validate the effectiveness of these two strategies and show promising potential for the proposed algorithm in solving EXDMOPs.
SWARM AND EVOLUTIONARY COMPUTATION
(2022)
Article
Computer Science, Artificial Intelligence
Liang Fan, Handing Wang
Summary: To accelerate performance estimation in neural architecture search, a surrogate-assisted evolutionary algorithm with network embedding (SAENAS-NE) is proposed. Unsupervised learning generates meaningful representation for each architecture, making architecture with similar structures closer in the embedding space, benefiting surrogate model training. A new environmental selection based on reference population and an infill criterion for balancing convergence and model uncertainty are introduced. Experimental results demonstrate the superiority of SAENAS-NE over other state-of-the-art algorithms.
COMPLEX & INTELLIGENT SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Yapei Wu, Xingguang Peng, Handing Wang, Yaochu Jin, Demin Xu
Summary: Many real-world optimization tasks suffer from noise, but current research on noise-tolerant algorithms is limited to low-dimensional problems. This article proposes a landscape-aware grouping method for cooperative coevolutionary algorithms to solve high-dimensional problems under noisy environments. Experimental results show that the proposed method is able to effectively identify interactive decision variables in the presence of noise.
IEEE TRANSACTIONS ON EVOLUTIONARY COMPUTATION
(2023)
Article
Computer Science, Artificial Intelligence
Junfeng Tang, Handing Wang, Lin Xiong
Summary: In preference-based multi-objective optimization, knee solutions are the implicit preferred promising solutions. However, finding knee solutions is difficult and computationally expensive. To address this issue, we propose a surrogate-assisted evolutionary multi-objective optimization algorithm that uses surrogate models to replace expensive evaluations. Experimental results show that our proposed algorithm outperforms state-of-the-art knee identification evolutionary algorithms on most test problems within a limited computational budget.
SWARM AND EVOLUTIONARY COMPUTATION
(2023)
Article
Computer Science, Artificial Intelligence
Nan Zheng, Handing Wang
Summary: For noisy bi-objective optimization problems, the algorithm is affected differently by noise in different stages of optimization. The proposed adaptive switch strategy enables the algorithm to adaptively switch among different noise treatments based on the noise impact. Additionally, data selection and model performance estimation methods are employed to enhance the denoising process, and reliable non-dominated solutions are selected as the final output. Experimental results demonstrate that the proposed algorithm is highly competitive for solving noisy bi-objective optimization problems.
SWARM AND EVOLUTIONARY COMPUTATION
(2023)
Article
Computer Science, Artificial Intelligence
Chao Li, Wen Yao, Handing Wang, Tingsong Jiang, Xiaoya Zhang
Summary: Due to the importance of security, adversarial attacks in deep learning, especially the black-box adversarial attack, which mimics real-world scenarios, have gained popularity. Query-based methods are commonly used for black-box attacks but suffer from needing excessive queries. To overcome this, a Bayesian evolutionary optimization (BEO) based black-box attack method using differential evolution is proposed, employing Gaussian processes model and adaptive acquisition functions. Experimental results show that this method can effectively generate high-quality adversarial examples using only 200 queries.
APPLIED SOFT COMPUTING
(2023)
Article
Automation & Control Systems
Zhening Liu, Handing Wang, Yaochu Jin
Summary: Offline data-driven multiobjective optimization problems are common in practice. To address the issue of error accumulation when using surrogate models for optimization, a new surrogate-assisted indicator-based evolutionary algorithm is proposed. This algorithm can select the appropriate type of surrogate models based on the error, and it performs well in practical problems.
IEEE TRANSACTIONS ON CYBERNETICS
(2023)
Article
Automation & Control Systems
Zhenshou Song, Handing Wang, Yaochu Jin
Summary: Expensive constrained optimization problems can be solved by evolutionary algorithms in conjunction with computationally cheap surrogates. However, existing methods neglect the differences between different surrogate models, resulting in unsatisfactory performance.
IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Jiliang Zhao, Handing Wang, Wen Yao, Wei Peng, Zhiqiang Gong
Summary: Thermal layout optimization problems are common in integrated circuit design, where optimizing the positions of electronic components is essential for achieving low temperatures. Surrogate models are used to reduce evaluation costs, but they often have large prediction errors in discrete decision spaces such as thermal layout problems. A deep neural network is proposed in this work to better approximate the relation between layout schemes and temperature fields, leading to an online deep surrogate-assisted optimization algorithm that effectively manages parameters for improved performance within limited computational budgets.
COMPLEX & INTELLIGENT SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Xudong Feng, Zhening Liu, Feng Wu, Handing Wang
Summary: Traditional engine cycle innovation is limited by human experiences, imagination, and currently available engine component performance expectations. In this study, a mixed variable multi-objective evolutionary optimization method is proposed for automatic engine cycle design. Through experimental research, new engine cycle solutions have been discovered that surpass the performance of known turbojet and turbofan engines.
COMPLEX & INTELLIGENT SYSTEMS
(2023)
Article
Engineering, Aerospace
Lei Han, Handing Wang, Shuo Wang
Summary: The study proposes a surrogate-assisted evolutionary algorithm that can find the optimal solution in expensive electronic component layout optimization problems. By combining local search and global search, and designing a restart strategy, the algorithm converges to the optimal solution more quickly.
SPACE: SCIENCE & TECHNOLOGY
(2022)
