Progressive Neural Architecture Search Method for Collaborative Optimization of Efficiency and Reliability

ZHANG Rui; WEI Xiaonan; SUN Chaoli

doi:10.12263/DZXB.20260222

您当前的位置：

首页 >

文章列表页 >

Progressive Neural Architecture Search Method for Collaborative Optimization of Efficiency and Reliability

PAPERS | 更新时间：2026-06-17

- Progressive Neural Architecture Search Method for Collaborative Optimization of Efficiency and Reliability
- ACTA ELECTRONICA SINICA Vol. 54, Issue 4, Pages: 1792-1805(2026)
- 作者机构：
  
  太原科技大学计算机科学与技术学院，山西太原 030024
- 作者简介：
- 基金信息：
  
  National Natural Science Foundation of China(62372319);Humanities and Social Sciences Research Project of the Ministry of Education(23YJCZH299);Key Research and Development Program of Shanxi Province(202302140601012);Special Program for Patent Transformation of Shanxi Province(20250021);Basic Research Program of Shanxi Province(202403021221142);Science and Technology Program of Shanxi Administration for Market Regulation(2025KJ019)
- DOI：10.12263/DZXB.20260222
  CLC： TP18;
- Received：13 March 2026，
  
  Accepted：08 April 2026，
  
  Published：25 April 2026
- 稿件说明：
移动端阅览
张睿, 魏晓楠, 孙超利. 协同优化效率与可靠性的递进式神经架构搜索方法[J]. 电子学报, 2026, 54(04): 1792-1805.

ZHANG Rui, WEI Xiaonan, SUN Chaoli. Progressive Neural Architecture Search Method for Collaborative Optimization of Efficiency and Reliability[J]. Acta Electronica Sinica, 2026, 54(04): 1792-1805.
张睿, 魏晓楠, 孙超利. 协同优化效率与可靠性的递进式神经架构搜索方法[J]. 电子学报, 2026, 54(04): 1792-1805. DOI：10.12263/DZXB.20260222

ZHANG Rui, WEI Xiaonan, SUN Chaoli. Progressive Neural Architecture Search Method for Collaborative Optimization of Efficiency and Reliability[J]. Acta Electronica Sinica, 2026, 54(04): 1792-1805. DOI：10.12263/DZXB.20260222

摘要

神经架构搜索作为自动化深度学习模型构建的核心技术，旨在搜索面向特定任务的最优网络结构，然而现有方法在以下方面存在不足：搜索空间中卷积与池化操作耦合，导致解空间冗余；高维编码采用整体优化方式，模块间协作不足，易陷入局部最优；评估策略依赖单一或少数指标，难以准确反映架构真实性能，易对搜索方向产生误导。上述问题相互交织，高潜力架构易被淘汰，搜索所得架构性能与最优值存在差距，限制了神经架构搜索（Neural Architecture Search， NAS）技术在资源受限及高可靠性场景中的应用。针对上述不足，本文提出一种递进式协同的神经架构搜索方法。该方法包含三个模块：在搜索空间层面，设计递进式表征降维可分离架构搜索空间，依据卷积与池化的功能特性将二者解耦，以缩小解空间规模并保留特征表征，为后续搜索与评估提供输入；在搜索策略层面，提出高维解耦式单元自适应搜索策略，将高维架构编码按卷积段、池化段、深度段进行分段解耦，对不同模块分别采用两点交叉、单点交叉及自适应变异等单元级定向遗传操作；在评估策略层面，构建多维低成本模型性能评估策略，从鲁棒性、各向同性、不确定性、规整性四个维度分别评估候选架构，并通过排名驱动的非线性几何融合机制将各维度分数整合为综合指标。上述三个模块依次衔接，形成递进式协同框架。实验结果表明，在NAS-Bench-201基准搜索空间上，所提方法在CIFAR-10、CIFAR-100及ImageNet16-120数据集上的肯德尔相关性系数分别为0.712 0、0.705 2与0.698 1，其中在CIFAR-10上较NASWOT方法提升20.13%。在工业焊缝缺陷检测与医疗APTOS-2019视网膜病变分级两项任务中，所得最优架构较现有无训练NAS方法平均精度分别提升约18.67%与6.12%，搜索耗时分别为227.97 s与559.34 s，推理延迟分别为0.41 ms与0.46 ms。该研究为构建高效、低成本且具备跨领域泛化能力的自动化模型设计技术提供了参考。

Abstract

Neural architecture search (NAS)

a core technology for automated deep learning model construction

aims to discover task-specific optimal network architectures. However

existing methods suffer from three critical limitations: coupled convolution and pooling operations in search spaces generate redundant solution spaces; holistically optimized high-dimensional encodings lack inter-module collaboration and easily fall into local optima; evaluation strategies relying on single or limited metrics fail to accurately reflect true architecture performance

biasing search directions. These intertwined issues eliminate high-potential architectures

create performance gaps between searched and optimal architectures

and restrict NAS deployment in resource-constrained and high-reliability scenarios. To address these deficiencies

this paper presents a progressive collaborative neural architecture search method. This method consists of three modules: a progressive representation dimensionality reduction separable architecture search space is designed at the search space level

where convolution and pooling are decoupled by their functional properties to reduce the solution space scale and retain feature representations

thus supplying input for subsequent search and evaluation. A high-dimensional decoupled unit-adaptive search strategy is proposed at the search strategy level

which decomposes high-dimensional architecture encodings into convolution segments

pooling segments and depth segments

and applies unit-level directed genetic operations including two-point crossover

single-point crossover and adaptive mutation to different modules separately. A multi-dimensional low-cost model performance evaluation strategy is constructed at the evaluation strategy level to assess candidate architectures from four dimensions

namely robustness

isotropy

uncertainty and regularity

and merges multi-dimensional scores into a comprehensive indicator through a rank-driven nonlinear geometric fusion mechanism. The three modules connect sequentially to form a progressive collaborative framework. Experimental results demonstrate that on the NAS-Bench-201 benchmark search space

the proposed method yields Kendall correlation coefficients of 0.712 0

0.705 2 and 0.698 1 on CIFAR-10

CIFAR-100 and ImageNet16-120 datasets respectively

with a 20.13% relative gain over NASWOT on CIFAR-10. For industrial weld defect detection and APTOS-2019 medical retinopathy grading

the derived optimal architecture boosts average accuracy by approximately 18.67% and 6.12% respectively over existing training-free NAS methods including NASWOT

ZiCo and MSTF-NAS

with search time of 227.97 s and 559.34 s

and inference latency of 0.41 ms and 0.46 ms correspondingly. This work provides a reference for developing efficient

low-cost automated model design techniques with strong cross-domain generalization.

关键词

Keywords

references

章晋睿 , 龙婷婷 , 张德宇 , 等 . 端智能推理加速技术综述 [J ] . 电子学报 , 2025 , 53 ( 4 ): 1063 - 1102 .

Zhang Jinrui , Long Tingting , Zhang Deyu , et al . On-device intelligence acceleration technologies: A survey [J ] . Acta Electronica Sinica , 2025 , 53 ( 4 ): 1063 - 1102 . (in Chinese)

Li Yangyang , Liu Guanlong , Shang Ronghua , et al . Meta knowledge assisted evolutionary neural architecture search [J ] . IEEE Transactions on Circuits and Systems for Video Technology , 2025 , 35 ( 10 ): 10225 - 10237 . DOI: 10.1109/TCSVT.2025.3565562 http://dx.doi.org/10.1109/TCSVT.2025.3565562

Liang Jing , Liu Genyue , Bi Ying , et al . Evolutionary neural architecture search for remote sensing image classification [J ] . IEEE Transactions on Neural Networks and Learning Systems , 2025 , 36 ( 10 ): 17886 - 17900 . DOI: 10.1109/tnnls.2025.3579517 http://dx.doi.org/10.1109/tnnls.2025.3579517

Zou Juan , Tong Jinghui , Xia Yizhang , et al . TS-ENAS: Two-stage evolution for cell-based network architecture search [J ] . Expert Systems with Applications , 2026 , 296 : 129150 . DOI: 10.1016/j.eswa.2025.129150 http://dx.doi.org/10.1016/j.eswa.2025.129150

蒋鹏程 , 薛羽 . 基于排序得分预测的演化神经架构搜索方法 [J ] . 计算机学报 , 2024 , 47 ( 11 ): 2522 - 2535 . DOI: 10.11897/SP.J.1016.2024.02522 http://dx.doi.org/10.11897/SP.J.1016.2024.02522

Jiang Pengcheng , Xue Yu . Evolutionary neural architecture search with predictor of ranking-based score [J ] . Chinese Journal of Computers , 2024 , 47 ( 11 ): 2522 - 2535 . (in Chinese) . DOI: 10.11897/SP.J.1016.2024.02522 http://dx.doi.org/10.11897/SP.J.1016.2024.02522

Yang Jiechao , Liu Yong , Wang Wei , et al . PATNAS: A path-based training-free neural architecture search [J ] . IEEE Transactions on Pattern Analysis and Machine Intelligence , 2025 , 47 ( 3 ): 1484 - 1500 . DOI: 10.1109/tpami.2024.3498035 http://dx.doi.org/10.1109/tpami.2024.3498035

Lopes V , Alexandre L A . Toward less constrained macro-neural architecture search [J ] . IEEE Transactions on Neural Networks and Learning Systems , 2025 , 36 ( 2 ): 2854 - 2868 . DOI: 10.1109/TNNLS.2023.3326648 http://dx.doi.org/10.1109/TNNLS.2023.3326648

Onzo B M , Xue Yu , Neri F . Surrogate-assisted evolutionary neural architecture search based on smart-block discovery [J ] . Expert Systems with Applications , 2025 , 277 : 127237 . DOI: 10.1016/j.eswa.2025.127237 http://dx.doi.org/10.1016/j.eswa.2025.127237

张睿 , 张鹏云 , 孙超利 . 基于多域融合及神经架构搜索的语音增强方法 [J ] . 通信学报 , 2024 , 45 ( 2 ): 225 - 239 . DOI: 10.11959/j.issn.1000-436x.2024018 http://dx.doi.org/10.11959/j.issn.1000-436x.2024018

Zhang Rui , Zhang Pengyun , Sun Chaoli . Speech enhancement method based on multi-domain fusion and neural architecture search [J ] . Journal on Communications , 2024 , 45 ( 2 ): 225 - 239 . (in Chinese) . DOI: 10.11959/j.issn.1000-436x.2024018 http://dx.doi.org/10.11959/j.issn.1000-436x.2024018

Wu Mengting , Lin H I , Tsai C W . A training-free neural architecture search algorithm based on search economics [J ] . IEEE Transactions on Evolutionary Computation , 2024 , 28 ( 2 ): 445 - 459 . DOI: 10.1109/TEVC.2023.3264533 http://dx.doi.org/10.1109/TEVC.2023.3264533

Yamasaki T , Wang Zhehui , Luo T , et al . RBFleX-NAS: Training-free neural architecture search using radial basis function kernel and hyperparameter detection [J ] . IEEE Transactions on Neural Networks and Learning Systems , 2025 , 36 ( 6 ): 10057 - 10071 . DOI: 10.1109/TNNLS.2025.3552693 http://dx.doi.org/10.1109/TNNLS.2025.3552693

杨乐 , 马永杰 , 平镐羽 , 等 . 角度修正和分级多种群的动态多目标进化算法 [J ] . 电子学报 , 2024 , 52 ( 9 ): 3278 - 3290 . DOI: 10.12263/DZXB.20221330 http://dx.doi.org/10.12263/DZXB.20221330

Yang Le , Ma Yongjie , Ping Gaoyu , et al . Dynamic multi-objective evolutionary algorithm based on angle correction and hierarchical multi-population [J ] . Acta Electronica Sinica , 2024 , 52 ( 9 ): 3278 - 3290 . (in Chinese) . DOI: 10.12263/DZXB.20221330 http://dx.doi.org/10.12263/DZXB.20221330

He Guang , Lu Xiaoli . Good point set and double attractors based-QPSO and application in portfolio with transaction fee and financing cost [J ] . Expert Systems with Applications , 2022 , 209 : 118339 . DOI: 10.1016/j.eswa.2022.118339 http://dx.doi.org/10.1016/j.eswa.2022.118339

Huang Ting , Tang Xiaoan , Zhao Shuangyao , et al . Linguistic information-based granular computing based on a tournament selection operator-guided PSO for supporting multi-attribute group decision-making with distributed linguistic preference relations [J ] . Information Sciences , 2022 , 610 : 488 - 507 . DOI: 10.1016/j.ins.2022.07.050 http://dx.doi.org/10.1016/j.ins.2022.07.050

Cai Xingyu , Huang Jiaji , Bian Yuchen , et al . Isotropy in the contextual embedding space: Clusters and manifolds [C/OL ] // Proceedings of the 9th International Conference on Learning Representations , 2021 . https://openreview.net/forum id=xYGNO86OWDH https://openreview.net/forumid=xYGNO86OWDH .

Karthik , Maggie , Dane S . APTOS 2019 blindness detection [EB/OL ] . [ 2026-03-13 ] . https://kaggle.com/competitions/aptos2019-blindness-detection https://kaggle.com/competitions/aptos2019-blindness-detection .

Dong Xuanyi , Yang Yi . NAS-bench-201: Extending the scope of reproducible neural architecture search [C/OL ] // Proceedings of the 8th International Conference on Learning Representations , 2020 . https://openreview.net/forum id=HJxyZkBKDr https://openreview.net/forumid=HJxyZkBKDr . DOI: 10.48550/arXiv.2001.00326 http://dx.doi.org/10.48550/arXiv.2001.00326

Lee N , Ajanthan T , Torr P H . Snip: Single-shot network pruning based on connection sensitivity [C/OL ] // Proceedings of the 7th International Conference on Learning Representations , 2019 . https://openreview.net/forum id=B1VZqjAcYX https://openreview.net/forumid=B1VZqjAcYX .

Wang Chaoqi , Zhang Guodong , Grosse R . Picking winning tickets before training by preserving gradient flow [C/OL ] // Proceedings of the 8th International Conference on Learning Representations , 2020 . https://openreview.net/forum id=SkgsACVKPH https://openreview.net/forumid=SkgsACVKPH . DOI: 10.48550/arXiv.2002.07376 http://dx.doi.org/10.48550/arXiv.2002.07376

Tanaka H , Kunin D , Yamins D L K , et al . Pruning neural networks without any data by iteratively conserving synaptic flow [C ] // Proceedings of the 34th International Conference on Neural Information Processing Systems . Red Hook : Curran Associates Inc. , 2020 : 535 .

Abdelfattah M S , Mehrotra A , Dudziak Ł , et al . Zero-cost proxies for lightweight NAS [C/OL ] // Proceedings of the 9th International Conference on Learning Representations , 2021 . https://openreview.net/forum id=0cmMMy8J5q https://openreview.net/forumid=0cmMMy8J5q .

Mellor J , Turner J , Storkey A , et al . Neural architecture search without training [C ] // Proceedings of the 38th International Conference on Machine Learning . PMLR , 2021 : 7588 - 7598 .

Zhang Zhihao , Jia Zhihao . GradSign: Model performance inference with theoretical insights [C/OL ] // Proceedings of the Tenth International Conference on Learning Representations , 2022 . https://openreview.net/forum id=HObMhrCeAAF https://openreview.net/forumid=HObMhrCeAAF .

Li Guihong , Yang Yuedong , Bhardwaj K , et al . ZiCo: Zero-shot NAS via inverse coefficient of variation on gradients [C/OL ] // Proceedings of the Eleventh International Conference on Learning Representations , 2023 . https://openreview.net/forum id=rwo-ls5GqGn https://openreview.net/forumid=rwo-ls5GqGn .

Huang Junhao , Xue Bing , Sun Yanan , et al . Efficient perturbation-aware distinguishing score for zero-shot neural architecture search [J ] . Applied Soft Computing , 2025 , 182 : 113447 . DOI: 10.1016/j.asoc.2025.113447 http://dx.doi.org/10.1016/j.asoc.2025.113447

Zhang Rui , Gao Meirong , Zhang Pengyun , et al . Research on an ultrasonic detection method for weld defects based on neural network architecture search [J ] . Measurement , 2023 , 221 : 113483 . DOI: 10.1016/j.measurement.2023.113483 http://dx.doi.org/10.1016/j.measurement.2023.113483

Wang Wenhai , Xie Enze , Li Xiang , et al . Pyramid vision transformer: A versatile backbone for dense prediction without convolutions [C ] // Proceedings of the 2021 IEEE/CVF International Conference on Computer Vision . Piscataway : IEEE , 2021 : 548 - 558 . DOI: 10.48550/arXiv.2102.12122 http://dx.doi.org/10.48550/arXiv.2102.12122

Dai Zihang , Liu Hanxiao , Le Q V , et al . CoAtNet: Marrying convolution and attention for all data sizes [C ] // Proceedings of the 35th International Conference on Neural Information Processing Systems . Red Hook : Curran Associates Inc. , 2021 : 303 . DOI: 10.48550/arXiv.2106.04803 http://dx.doi.org/10.48550/arXiv.2106.04803

Li Yanyu , Yuan Geng , Wen Yang , et al . EfficientFormer: Vision transformers at MobileNet speed [C ] // Proceedings of the 36th International Conference on Neural Information Processing Systems . Red Hook : Curran Associates Inc. , 2022 : 940 . DOI: 10.52202/068431-0940 http://dx.doi.org/10.52202/068431-0940

Mehta S , Rastegari M . Separable self-attention for mobile vision transformers [J ] . Transactions on Machine Learning Research , 2023 , 2023 .

Woo S , Debnath S , Hu Ronghang , et al . ConvNeXt V2: Co-designing and scaling ConvNets with masked autoencoders [C ] // Proceedings of the 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition . Piscataway : IEEE , 2023 : 16133 - 16142 . DOI: 10.1109/cvpr52729.2023.01548 http://dx.doi.org/10.1109/cvpr52729.2023.01548

Wang Wenhai , Xie Enze , Li Xiang , et al . PVT v2: Improved baselines with pyramid vision transformer [J ] . Computational Visual Media , 2022 , 8 ( 3 ): 415 - 424 . DOI: 10.1007/s41095-022-0274-8 http://dx.doi.org/10.1007/s41095-022-0274-8

Yu Weihao , Zhou Pan , Yan Shuicheng , et al . InceptionNeXt: When inception meets ConvNeXt [C ] // Proceedings of the 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition . Piscataway : IEEE , 2024 : 5672 - 5683 . DOI: 10.1109/cvpr52733.2024.00542 http://dx.doi.org/10.1109/cvpr52733.2024.00542

Lee S , Choi J , Kim H J . EfficientViM: Efficient vision mamba with hidden state mixer based state space duality [C ] // Proceedings of the 2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition . Piscataway : IEEE , 2025 : 14923 - 14933 . DOI: 10.1109/cvpr52734.2025.01390 http://dx.doi.org/10.1109/cvpr52734.2025.01390

Fan Qihang , Huang Huaibo , Ai Yuaang , et al . Rectifying magnitude neglect in linear attention [C ] // Proceedings of the IEEE/CVF International Conference on Computer Vision . Piscataway : IEEE , 2025 : 21505 - 21514 . DOI: 10.1109/iccv51701.2025.01997 http://dx.doi.org/10.1109/iccv51701.2025.01997

Shi Chaokun , Hao Yuexing , Li Gongyan , et al . EBNAS: Efficient binary network design for image classification via neural architecture search [J ] . Engineering Applications of Artificial Intelligence , 2023 , 120 : 105845 . DOI: 10.1016/j.engappai.2023.105845 http://dx.doi.org/10.1016/j.engappai.2023.105845

Ali M J , Moalic L , Essaid M , et al . Evolutionary neural architecture search for 2D and 3D medical image classification [C ] // Proceedings of the 24th International Conference on Computational Science . Cham : Springer , 2024 : 131 - 146 . DOI: 10.1007/978-3-031-63751-3_9 http://dx.doi.org/10.1007/978-3-031-63751-3_9

Lin Mingzhuo , Luo Jianping . Per-architecture training-free metric optimization for neural architecture search [C ] // The Thirty-Ninth Annual Conference on Neural Information Processing Systems , 2025 . https://openreview.net/forum id=mVh0lIsdUl https://openreview.net/forumid=mVh0lIsdUl .

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

A Multi-Task Evolutionary Algorithm for Multi-Objective Vehicle Routing Problems

Location Mapping and Correlation Assessment Based Evolutionary Multi-Task Optimization Algorithm in the Internet of Vehicles

MaOEA/A2R: A Many-Objective Evolutionary Algorithm Based on A2R Dominance Relation

Constrained Dropout Surrogate-Assisted Evolutionary Algorithm for Expensive Many-Objective Problems

A Full-Scene Detection Algorithm for Underground Parking Spaces Fusing Vision and Geometric Reasoning

Related Author

CAO Longhao

LU Yuming

GUO Xin

AI Yihao

CHEN Hao

JIE Lilin

SHEN Jun-jie

PENG Jiang

Related Institution

Zhuhai BOE MLED Technology Co., Ltd.

School of Instrument Science and Optoelectronic Engineering, Nanchang Hangkong University

School of Information Engineering, Nanchang Hangkong University

Beijing Anqisheng Technology Co., Ltd.

School of Aeronautical Manufacturing and Mechanical Engineering, Nanchang Hangkong University

⁰