基于多核数据合成的离线小数据驱动的进化算法

doi:10.3785/j.issn.1008-973X.2025.02.006

浙江大学学报(工学版)

2025, Vol. 59

Issue (2): 278-288 DOI: 10.3785/j.issn.1008-973X.2025.02.006

计算机技术

基于多核数据合成的离线小数据驱动的进化算法

李二超(

),刘昀

兰州理工大学电气工程与信息工程学院，甘肃兰州 730050

Offline small data-driven evolutionary algorithm based on multi-kernel data synthesis

Erchao LI(

),Yun LIU

College of Electrical and Information Engineering, Lanzhou University of Technology, Lanzhou 730050, China

全文: PDF(838 KB) HTML

摘要：

为了增强离线数据驱动的进化算法在小数据情景中的表现, 削弱代理模型对数据集规模的依赖, 提出基于多核数据合成的离线小数据驱动的进化算法(DDEA-MKDS). 考虑到代理模型易因小数据陷入过拟合, 通过经验公式与遍历法找出针对离线数据集的最优隐含层节点数，以简化模型结构. 为了弥补数据量的不足, 训练了3个不同核函数的径向基网络生成合成数据, 通过轮盘赌法选择其中的部分数据与原数据集合并, 使用新数据集训练代理模型. 将DDEA-MKDS与其他5种流行的离线数据驱动的进化算法在6个单目标基准测试问题上进行对比, 实验结果表明, 所提算法在数据量极小的条件下能够取得良好的效果, 寻优效率显著优于其他算法.

关键词： 离线数据驱动; 进化算法; 小数据; 代理模型; 隐含层节点; 合成数据

Abstract:

An offline data-driven evolutionary algorithm based on multi-kernel data synthesis (DDEA-MKDS) was proposed to enhance the performance of such algorithms in small data scenarios and weaken the dependence of surrogate model on data set size. The empirical formula and traversal method were used to calculate the optimal number of hidden layer nodes for the offline data set in order to simplify model structure by considering that the surrogate model is prone to overfitting due to small data. Three radial basis networks with different kernel functions were trained to generate synthetic data in order to make up for the lack of data. A part of synthetic data was selected by roulette to combine with original data, and new data set was used to train the surrogate model. The experimental results showed that DDEA-MKDS had good performance under the condition of small data by comparing with five states of the art offline data-driven evolutionary algorithms on six single objective benchmark problems, and its efficiency was significantly better than other algorithms.

Key words: offline data-driven evolutionary algorithm small data surrogate model hidden layer node synthetic data

收稿日期: 2023-12-13 出版日期: 2025-02-11

CLC:

TP 181

基金资助: 国家自然科学基金资助项目（62063019）；甘肃省自然科学基金资助项目（24JRRA173，22JR5RA241）.

作者简介: 李二超（1980—），男，教授，从事人工智能、进化计算的研究. https：//orcid.org/0000-0001-7050-072X.E-mail：lecstarr@163.com

	服务
	把本文推荐给朋友
	加入引用管理器
	E-mail Alert
	作者相关文章
	李二超
	刘昀

引用本文:

李二超,刘昀. 基于多核数据合成的离线小数据驱动的进化算法[J]. 浙江大学学报(工学版), 2025, 59(2): 278-288.

Erchao LI,Yun LIU. Offline small data-driven evolutionary algorithm based on multi-kernel data synthesis. Journal of ZheJiang University (Engineering Science), 2025, 59(2): 278-288.

链接本文:

https://www.zjujournals.com/eng/CN/10.3785/j.issn.1008-973X.2025.02.006 或 https://www.zjujournals.com/eng/CN/Y2025/V59/I2/278

图 1 径向基函数网络的结构

图 2 DDEA-MKDS的总流程图

问题	d	DDEA-MKDS	TT-DDEA	SRK-DDEA	CC-DDEA	CL-DDEA	DDEA-SE
Ellipsoid	10	0.79±0.73	4.52±6.36(+)	1.91±2.52(≈)	1.91±1.12(+)	114.41±97.15(+)	2.81±1.38(+)
	30	5.58±2.99	18.68±14.99(+)	11.40±11.59(+)	9.05±5.60(+)	451.17±394.59(+)	48.57±21.00(+)
	50	27.88±9.13	110.19±84.30(+)	23.99±14.62(?)	32.19±12.61(+)	1181.39±1035.79(+)	195.13±57.46(+)
	100	252.72±52.58	1380.95±1237.13(+)	269.84±54.97(≈)	49.60±22.66(?)	4908.28±5817.24(+)	1399.45±377.21(+)
Rosenbrock	10	16.22±7.00	20.84±10.33(≈)	23.90±12.02(+)	31.33±15.88(+)	941.78±796.76(+)	31.81±14.69(+)
	30	39.82±9.80	57.50±25.14(+)	65.48±29.53(+)	61.12±16.51(+)	1170.28±1082.81(+)	80.95±28.13(+)
	50	69.24±18.95	125.76±37.03(+)	85.74±20.65(≈)	104.91±51.58(+)	1298.44±1265.39(+)	217.48±61.36(+)
	100	209.61±36.64	373.37±68.49(+)	189.10±24.13(≈)	178.94±42.90(≈)	1907.06±1789.96(+)	714.23±121.78(+)
Ackley	10	5.78±1.73	9.71±4.75(+)	6.42±1.90(≈)	7.57±2.24(+)	10.27±4.71(≈)	6.35±1.63(≈)
	30	4.25±0.69	6.91±1.98(+)	6.10±2.17(+)	8.19±4.71(+)	16.19±2.45(+)	8.96±1.35(+)
	50	4.73±0.47	7.63±0.81(+)	5.94±1.65(≈)	5.73±1.98(≈)	11.51±3.09(+)	10.25±0.86(+)
	100	7.20±0.57	11.02±2.01(+)	7.28±0.60(≈)	4.95±0.85(?)	13.01±4.70(+)	11.59±0.52(+)
Levy	10	1.60±0.36	3.19±3.00(≈)	2.30±1.22(≈)	2.48±2.42(≈)	18.56±9.85(+)	2.95±2.53(≈)
	30	3.57±0.76	10.18±10.03(+)	4.39±1.33(≈)	4.02±0.87(≈)	51.71±39.59(+)	8.62±3.44(+)
	50	5.94±0.77	16.57±11.72(+)	6.01±1.06(≈)	10.30±4.54(+)	109.68±64.49(+)	20.91±6.25(+)
	100	13.70±3.59	106.97±66.70(+)	13.88±3.64(≈)	11.03±1.97(≈)	154.60±79.65(+)	73.42±17.79(+)
Griewank	10	1.21±0.28	3.51±3.64(+)	1.12±0.11(≈)	1.19±0.21(≈)	26.32±25.28(+)	2.26±0.54(+)
	30	3.32±1.00	5.72±3.12(+)	1.47±0.14(?)	1.70±0.33(?)	96.19±122.65(+)	12.34±4.19(+)
	50	6.24±1.81	9.23±6.83(≈)	3.04±0.51(?)	2.21±0.40(?)	136.93±114.01(+)	22.87±6.96(+)
	100	21.39±4.12	61.68±30.82(+)	18.56±3.96(≈)	3.61±0.74(?)	148.10±168.05(+)	90.66±20.69(+)
Rastrigin	10	42.23±26.46	58.47±28.90(≈)	64.47±35.35(+)	89.87±37.44(+)	138.64±34.53(+)	76.92±26.52(≈)
	30	95.41±61.47	242.85±162.46(+)	152.73±104.34(≈)	195.63±60.41(+)	322.84±68.42(+)	219.82±36.95(+)
	50	184.75±61.41	397.02±89.28(+)	245.34±101.93(+)	280.60±89.96(+)	570.56±47.71(+)	431.38±57.44(+)
	100	692.34±122.68	1061.33±190.85(+)	697.66±128.78(≈)	319.72±271.23(?)	1307.19±388.38(+)	953.94±51.30(+)
+/≈/?		NA	20/4/0	6/15/3	12/6/6	23/1/0	21/3/0
Friedman Rank		1.63	4.04	2.40	2.40	6.00	4.54

表 1 DDEA-MKDS与各对比算法在6个测试问题上的优化结果

图 3 DDEA-MKDS与各对比算法在Ellipsoid与Rastrigin问题上的收敛曲线

图 4 DDEA-MKDS与各对比算法在所有问题上的平均运行时间

表 2 DDEA-MKDS与各变体算法在Ellipsoid与Rastrigin问题上的优化结果

图 5 DDEA-MKDS与各变体算法在所有问题上的平均运行时间

表 3 使用不同核函数时DDEA-MKDS在Ellipsoid与Rastrigin问题上的优化结果

表 4 使用不同数量核函数时DDEA-MKDS的平均运行时间

表 5 dm取不同值时DDEA-MKDS在Ellipsoid与Rastrigin问题上的优化结果

图 6 dm取不同值时DDEA-MKDS的优化结果的变化趋势

1	JIN Y, WANG H, CHUGH T, et al Data-driven evolutionary optimization: an overview and case studies[J]. IEEE Transactions on Evolutionary Computation, 2019, 23 (3): 442- 458 doi: 10.1109/TEVC.2018.2869001
2	CHEN R, HE C, JIN Y, et al Model-based evolutionary algorithms: a short survey[J]. Complex and Intelligent Systems, 2018, 4 (4): 283- 292 doi: 10.1007/s40747-018-0080-1
3	HE C, TIAN Y, WANG H, et al A repository of real-world datasets for data-driven evolutionary multiobjective optimization[J]. Complex and Intelligent Systems, 2020, 6 (1): 189- 197 doi: 10.1007/s40747-019-00126-2
4	WANG H, JIN Y, JANSEN J O Data-driven surrogate-assisted multiobjective evolutionary optimization of a trauma system[J]. IEEE Transactions on Evolutionary Computation, 2016, 20 (6): 939- 952 doi: 10.1109/TEVC.2016.2555315
5	GUO D, CHAI T, DING J, et al. Small data driven evolutionary multi-objective optimization of fused magnesium furnaces [C]// IEEE Symposium Series on Computational Intelligence . Athens: IEEE, 2016: 1-8.
6	黄鹏飞. 离线数据驱动的进化优化研究[D]. 西安: 西安电子科技大学, 2021: 15-18, 46-47. HUANG Pengfei. A study on offline data-driven evolutionary optimization [D]. Xi’an: Xidian University, 2021: 15-18, 46-47.
7	HUANG P, WANG H, MA W. Stochastic ranking for offline data-driven evolutionary optimization using radial basis function networks with multiple kernels [C]// IEEE Symposium Series on Computational Intelligence . Xiamen: IEEE, 2019: 2050-2057.
8	CHENG R, JIN Y, NARUKAWA K, et al A multiobjective evolutionary algorithm using gaussian process-based inverse modeling[J]. IEEE Transactions on Evolutionary Computation, 2015, 19 (6): 838- 856 doi: 10.1109/TEVC.2015.2395073
9	梁正平, 黄锡均, 李燊钿, 等基于剪枝堆栈泛化的离线数据驱动进化优化[J]. 自动化学报, 2023, 49 (6): 1306- 1325 LIANG Zhengping, HUANG Xijun, LI Shentian, et al Offline data driven evolutionary optimization based on pruning stacked generalization[J]. Acta Automatica Sinica, 2023, 49 (6): 1306- 1325
10	ZHOU Z, ONG Y S, NGUYEN M H, et al. A study on polynomial regression and gaussian process global surrogate model in hierarchical surrogate-assisted evolutionary algorithm [C]// IEEE Congress on Evolutionary Computation . Edinburgh: IEEE, 2005: 2832-2839.
11	HUANG P, WANG H, JIN Y Offline data-driven evolutionary optimization based on tri-training[J]. Swarm and Evolutionary Computation, 2021, 60: 100800 doi: 10.1016/j.swevo.2020.100800
12	CHUGH T, CHAKRABORTI N, SINDHYA K, et al A data-driven surrogate-assisted evolutionary algorithm applied to a many-objective blast furnace optimization problem[J]. Materials and Manufacturing Processes, 2017, 32 (10): 1172- 1178 doi: 10.1080/10426914.2016.1269923
13	LI J, ZHAN Z, ZHANG J Evolutionary computation for expensive optimization: a survey[J]. Machine Intelligence Research, 2022, 19 (1): 3- 23 doi: 10.1007/s11633-022-1317-4
14	CHUGH T, JIN Y, MIETTINEN K, et al A surrogate-assisted reference vector guided evolutionary algorithm for computationally expensive many-objective optimization[J]. IEEE Transactions on Evolutionary Computation, 2016, 22 (1): 129- 142
15	LIM D, ONG Y S, JIN Y, et al. A study on metamodeling techniques, ensembles, and multi-surrogates in evolutionary computation [C]// Proceedings of the 9th Annual Conference on Genetic and Evolutionary Computation . London: ACM, 2007: 1288-1295.
16	MEY A, LOOG M Improved generalization in semi-supervised learning: a survey of theoretical results[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023, 45 (4): 4747- 4767 doi: 10.1109/TPAMI.2022.3198175
17	GÖNEN M, ALPAYDIN E Multiple kernel learning algorithms[J]. The Journal of Machine Learning Research, 2011, 12 (7): 2211- 2268
18	MA X, LI X, ZHANG Q, et al A survey on cooperative co-evolutionary algorithms[J]. IEEE Transactions on Evolutionary Computation, 2019, 23 (3): 421- 441 doi: 10.1109/TEVC.2018.2868770
19	WANG H, JIN Y, SUN C, et al Offline data-driven evolutionary optimization using selective surrogate ensembles[J]. IEEE Transactions on Evolutionary Computation, 2018, 23 (2): 203- 216
20	ZHOU Z, LI M Tri-training: exploiting unlabeled data using three classifiers[J]. IEEE Transactions on Knowledge and Data Engineering, 2005, 17 (11): 1529- 1541 doi: 10.1109/TKDE.2005.186
21	GOSAIN A, SACHDEVA K Materialized view selection for query performance enhancement using stochastic ranking based cuckoo search algorithm[J]. International Journal of Reliability, Quality and Safety Engineering, 2020, 27 (3): 2050008 doi: 10.1142/S0218539320500084
22	GONG Y, ZHONG Y, HUANG H. Offline data-driven optimization at scale: a cooperative coevolutionary approach [EB/OL]. (2023-12-04)[2024-01-30]. https://doi.org/10.1109/TEVC.2023.3338693.
23	HUANG H, GONG Y Contrastive learning: an alternative surrogate for offline data-driven evolutionary computation[J]. IEEE Transactions on Evolutionary Computation, 2023, 27 (2): 370- 384 doi: 10.1109/TEVC.2022.3170638
24	WANG R, ZHU F, ZHANG X, et al Training with scaled logits to alleviate class-level over-fitting in few-shot learning[J]. Neurocomputing, 2023, 522: 142- 151 doi: 10.1016/j.neucom.2022.12.011
25	ROHLOFF C T, KOHLI N, CHUNG S The impact of functional form complexity on model overfitting for nonlinear mixed-effects models[J]. Multivariate Behavioral Research, 2023, 58 (4): 723- 742 doi: 10.1080/00273171.2022.2119360
26	RASTEGAR R, HARIRI A A step forward in studying the compact genetic algorithm[J]. Evolutionary Computation, 2006, 14 (3): 277- 289 doi: 10.1162/evco.2006.14.3.277
27	王嵘冰, 徐红艳, 李波, 等 BP神经网络隐含层节点数确定方法研究[J]. 计算机技术与发展, 2018, 28 (4): 31- 35 WANG Rongbing, XU Hongyan, LI Bo, et al Research on method of determining hidden layer nodes in BP neural network[J]. Computer Technology and Development, 2018, 28 (4): 31- 35 doi: 10.3969/j.issn.1673-629X.2018.04.007
28	MAO Y, LIU C, XIAO D, et al Study of the magnetic properties of haematite based on spectroscopy and the IPSO-ELM neural network[J]. Journal of Sensors, 2018, 2018 (1): 1- 9
29	MIN M, CHEN X, LEI Y, et al A novel kernel-based extreme learning machine with incremental hidden layer nodes[J]. IFAC PapersOnLine, 2020, 53 (2): 11836- 11841 doi: 10.1016/j.ifacol.2020.12.695
30	TAO L, CAO T, WANG Q, et al Distribution adaptation and classification framework based on multiple kernel learning for motor imagery BCI illiteracy[J]. Sensors, 2022, 22 (17): 6572 doi: 10.3390/s22176572
31	PRICE S R, ANDERSEN D T, HAVENS T C, et al Kernel matrix-based heuristic multiple kernel learning[J]. Mathematics, 2022, 10 (12): 2026 doi: 10.3390/math10122026

[1]	吴菲,陈嘉诚,王万良. 基于并行计算的计算智能综述[J]. 浙江大学学报(工学版), 2025, 59(1): 27-38.
[2]	喻伯平,李高华,谢亮,王福新. 基于代理模型的旋翼翼型动态失速优化设计[J]. 浙江大学学报(工学版), 2020, 54(4): 833-842.
[3]	张玄武, 郑耀, 杨波威, 张继发. 基于级联前向网络的翼型优化设计[J]. 浙江大学学报(工学版), 2017, 51(7): 1405-1411.
[4]	过晓芳,王宇平,代才. 新的混合分解高维多目标进化算法[J]. 浙江大学学报(工学版), 2016, 50(7): 1313-1321.
[5]	王云, 冯毅雄, 谭建荣, 高一聪. 柔性作业车间分批调度多目标优化方法[J]. J4, 2011, 45(4): 719-726.
[6]	赵朋, 傅建中, 李阳, 崔树标. 基于代理模型的注射参数迭代优化方法[J]. J4, 2011, 45(2): 197-200.
[7]	刘林, 俞小莉, 胡军强, 等. 用于气动发动机设计的非支配排序遗传算法的改进[J]. J4, 2009, 43(5): 907-910.
[8]	刘林俞小莉胡军强陈平录. 基于Pareto Frontier方法的气动发动机设计[J]. J4, 2009, 43(1): 123-127.
[9]	高一聪谭建荣冯毅雄魏喆. 复杂机械产品性能意图优化研究[J]. J4, 2008, 42(9): 1549-1553.

Viewed

Full text

Abstract

Cited

Shared

Discussed