A classification surrogate-assisted multi-objective evolutionary algorithm for expensive optimization

Li, Jinglu; Wang, Peng; Dong, Huachao; Shen, Jiangtao; Chen, Caihua

doi:10.1016/j.knosys.2022.108416

Cited by 27 publications

(8 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It is well-known that the exact fitness values are unnecessary during selection if the binary relation between solutions MOEA/D-EGO [25] K-RVEA [26] ParEGO [31] ... CSEA [47] CSA-MOEA [48] MCEA/D [49] ... RCPS [52] θ-DEA-DP [54] REMO [55] ... can be found. Coincidentally, grid location is one of the ways that can reflect this relation of solutions.…”

Section: B Grid Definition and Grid Dominancementioning

confidence: 99%

“…Some specific parameters about them are shown in Table I. Apart from that, GCS-PSO is compared with seven state-of-the-art SAEAs, including three regression-based SAEAs (K-RVEA [26], HeE-MOEA [35], and EDN-ARMOEA [40]) and four classification-based SAEAs (CSA-MOEA [48], CSEA [47], MCEA/D [49], and REMO [55]). Each algorithm is conducted on three-objective DTLZ1-DTLZ5 [68] and WFG1-WFG9 [69] benchmark problems with different dimensions of decision variables D = 20, 40, 60, and 80.…”

Section: A Parameter Settingsmentioning

confidence: 99%

“…In the classification-based surrogate-assisted evolutionary algorithm (CSEA) [47], solutions in the population were classified into two categories by an FNN. Moreover, the classification surrogate-assisted MOEA (CSA-MOEA) [48] partitioned the population into several clusters by reference vectors, and only solutions with good performance in each cluster were labeled as '1'. Offspring are allowed to be preserved only when labeled as '1' using a classification tree.…”

Section: Introductionmentioning

confidence: 99%

“…This makes both the rank classification-based and the relation classification-based SAEAs require additional strategies to maintain diversity. For example, CSA-MOEA [48] used the promoting method, while θ-DEA-DP [54] used predefined uniform reference vectors to maintain diversity. Moreover, as the diversity information is difficult to be considered in existing classification-based SAEAs, most of them still use complex network models as classifiers.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Grid Classification-Based Surrogate-Assisted Particle Swarm Optimization for Expensive Multiobjective Optimization

Yang,

Zhan,

Liu

et al. 2023

IEEE Trans. Evol. Computat.

View full text Add to dashboard Cite

Surrogate-assisted evolutionary algorithms (SAE-As), mainly including regression-based SAEAs and classification-based SAEAs, are promising for solving expensive multiobjective optimization problems (EMOPs). Regression-based SAEAs usually use complex regression models to approximate the fitness evaluation, which will suffer from high training costs to obtain a fine-accuracy surrogate. In contrast, classification-based SAEAs can achieve solution selection via coarse binary relations predicted by classifiers, thus avoiding high requirements in prediction accuracy and training costs. However, most of the binary relations in existing classification-based SAEAs mainly only involve convergence comparison whereas diversity maintenance is neglected. Considering the capacity of the grid technique in maintaining both convergence and diversity, we propose a new classification method called grid classification to discretize the objective space into grids and train a lightweight grid classification-based surrogate (GCS), for which low training costs are needed. The GCS can evaluate the solution performance in terms of both convergence and diversity simultaneously according to the predicted grid locations, which opens up a new field for follow-up research on classification-based SAEAs. Following this, a GCS-assisted particle swarm optimization algorithm is proposed for tackling EMOPs. Experimental results on widely-used benchmark problems (including high-dimensional EMOPs) and a 222-high-dimensional real-world application problem show its competitiveness in terms of both optimization performance and computational cost.

show abstract

Section: B Grid Definition and Grid Dominancementioning

confidence: 99%

Section: A Parameter Settingsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Grid Classification-Based Surrogate-Assisted Particle Swarm Optimization for Expensive Multiobjective Optimization

Yang,

Zhan,

Liu

et al. 2023

IEEE Trans. Evol. Computat.

View full text Add to dashboard Cite

show abstract

“…Therein, most of the operations are carried out at the level of fast surrogates, with costly EM simulations only executed occasionally, to validate the designs produced using the metamodels or to obtain the data necessary for model refinement. Among the two major classes of surrogate modelling methods, the physics-based ones are more often used for local search purposes (space mapping 43 , response correction 44 – 46 ), whereas data-driven models (kriging 47 , Gaussian process regression, GPR 48 , artificial neural networks 49 – 51 , support vector regression 52 , polynomial chaos expansion 53 ) are perceived as more generic, and suitable for global and multi-criterial design 54 – 56 , as well as uncertainty quantification 57 – 60 . Related methods include machine learning techniques 61 – 63 , as well as surrogate-assisted frameworks involving variable-resolution models (two-level GPR 64 , co-kriging 65 ).…”

Section: Introductionmentioning

confidence: 99%

Knowledge-based expedited parameter tuning of microwave passives by means of design requirement management and variable-resolution EM simulations

Kozieł

Pietrenko‐Dabrowska

Raef

2023

Sci Rep

View full text Add to dashboard Cite

The importance of numerical optimization techniques has been continually growing in the design of microwave components over the recent years. Although reasonable initial designs can be obtained using circuit theory tools, precise parameter tuning is still necessary to account for effects such as electromagnetic (EM) cross coupling or radiation losses. EM-driven design closure is most often realized using gradient-based procedures, which are generally reliable as long as the initial design is sufficiently close to the optimum one. Otherwise, the search process may end up in a local optimum that is of insufficient quality. Furthermore, simulation-based optimization incurs considerable computational expenses, which are often impractically high. This paper proposes a novel parameter tuning procedure, combining a recently reported design specification management scheme, and variable-resolution EM models. The former allows for iteration-based automated modification of the design goals to make them accessible in every step of the search process, thereby improving its immunity to poor starting points. The knowledge-based procedure for the adjustment of the simulation model fidelity is based on the convergence status of the algorithm and discrepancy between the current and the original performance specifications. Due to using lower-resolution EM simulations in early phase of the optimization run, considerable CPU savings can be achieved, which are up to 60 percent over the gradient-based search employing design specifications management and numerical derivatives. Meanwhile, as demonstrated using three microstrip circuits, the computational speedup is obtained without design quality degradation.

show abstract

A Surrogate-Assisted Multi-objective Evolutionary Algorithm Guided by Hybrid Reference Points

Li,

Zhang,

Wang

et al. 2024

Lecture Notes in Computer Science

View full text Add to dashboard Cite

A classification surrogate-assisted multi-objective evolutionary algorithm for expensive optimization

Cited by 27 publications

References 42 publications

Grid Classification-Based Surrogate-Assisted Particle Swarm Optimization for Expensive Multiobjective Optimization

Grid Classification-Based Surrogate-Assisted Particle Swarm Optimization for Expensive Multiobjective Optimization

Knowledge-based expedited parameter tuning of microwave passives by means of design requirement management and variable-resolution EM simulations

A Surrogate-Assisted Multi-objective Evolutionary Algorithm Guided by Hybrid Reference Points

Contact Info

Product

Resources

About