Maintained by Difan Deng and Marius Lindauer.
The following list considers papers related to neural architecture search. It is by no means complete. If you miss a paper on the list, please let us know.
Please note that although NAS methods steadily improve, the quality of empirical evaluations in this field are still lagging behind compared to other areas in machine learning, AI and optimization. We would therefore like to share some best practices for empirical evaluations of NAS methods, which we believe will facilitate sustained and measurable progress in the field. If you are interested in a teaser, please read our blog post or directly jump to our checklist.
Transformers have gained increasing popularity in different domains. For a comprehensive list of papers focusing on Neural Architecture Search for Transformer-Based spaces, the awesome-transformer-search repo is all you need.
2025
4089.
Kazadi, Joël
Exploring Optimal Neural Network Architectures: What benefits does Reinforcement Learning offer? Bachelor Thesis
2025.
@bachelorthesis{unknownc,
title = {Exploring Optimal Neural Network Architectures: What benefits does Reinforcement Learning offer?},
author = {Joël Kazadi},
url = {https://www.researchgate.net/publication/392493431_Exploring_Optimal_Neural_Network_Architectures_What_benefits_does_Reinforcement_Learning_offer},
doi = {10.13140/RG.2.2.17572.18564},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
keywords = {},
pubstate = {published},
tppubtype = {bachelorthesis}
}
4088.
Yang, Yu; Wang, Siqi; Zhang, Gan; Wang, Qifu; Qin, Yao; Zhai, Dandan; Yang, Zhiqing; Li, Peng
GA-OMTL: Genetic algorithm optimization for multi-task neural architecture search in NIR spectroscopy Journal Article
In: Expert Systems with Applications, vol. 290, pp. 128517, 2025, ISSN: 0957-4174.
@article{YANG2025128517,
title = {GA-OMTL: Genetic algorithm optimization for multi-task neural architecture search in NIR spectroscopy},
author = {Yu Yang and Siqi Wang and Gan Zhang and Qifu Wang and Yao Qin and Dandan Zhai and Zhiqing Yang and Peng Li},
url = {https://www.sciencedirect.com/science/article/pii/S0957417425021360},
doi = {https://doi.org/10.1016/j.eswa.2025.128517},
issn = {0957-4174},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
journal = {Expert Systems with Applications},
volume = {290},
pages = {128517},
abstract = {Near-infrared (NIR) spectroscopy analysis based on deep learning has emerged as a powerful tool for the quality assessment of food and agricultural products. However, most existing multi-task deep learning (MTL) network architectures rely on manual design, struggle to efficiently adapt to diverse datasets, and often neglect the interaction of task-specific (private) features across tasks. To address these challenges, this study proposes a genetic algorithm (GA)-optimized MTL model, termed GA-OMTL, which integrates the strengths of neural architecture search and GA for multi-task prediction of spectral data. The model enhances both feature extraction and task-specific feature interaction by incorporating searchable components such as residual modules (Resblock), batch normalization (BN) layers, Squeeze-and-Excitation (SE) modules, and feature interaction modules. The effectiveness of GA-OMTL was validated using two datasets: American ginseng and wheat flour. In the prediction of protopanaxatriol-type ginsenosides (PPT) and protopanaxadiol-type ginsenosides (PPD) in American ginseng, the R2, RMSE, and RPD values achieved by GA-OMTL were 0.93, 0.70 mg/g, and 3.83 (PPT), and 0.98, 2.03 mg/g, and 7.16 (PPD), respectively. For the prediction of protein and moisture content in wheat flour, the R2, RMSE, and RPD values were 0.99, 0.29 mg/g, and 8.22 (protein), and 0.97, 0.22 mg/g, and 5.67 (moisture), respectively. The experimental results demonstrate that GA-OMTL outperforms three comparison methods in prediction accuracy, highlighting its potential for complex spectral tasks and confirming the practicality and robustness of the proposed model.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Near-infrared (NIR) spectroscopy analysis based on deep learning has emerged as a powerful tool for the quality assessment of food and agricultural products. However, most existing multi-task deep learning (MTL) network architectures rely on manual design, struggle to efficiently adapt to diverse datasets, and often neglect the interaction of task-specific (private) features across tasks. To address these challenges, this study proposes a genetic algorithm (GA)-optimized MTL model, termed GA-OMTL, which integrates the strengths of neural architecture search and GA for multi-task prediction of spectral data. The model enhances both feature extraction and task-specific feature interaction by incorporating searchable components such as residual modules (Resblock), batch normalization (BN) layers, Squeeze-and-Excitation (SE) modules, and feature interaction modules. The effectiveness of GA-OMTL was validated using two datasets: American ginseng and wheat flour. In the prediction of protopanaxatriol-type ginsenosides (PPT) and protopanaxadiol-type ginsenosides (PPD) in American ginseng, the R2, RMSE, and RPD values achieved by GA-OMTL were 0.93, 0.70 mg/g, and 3.83 (PPT), and 0.98, 2.03 mg/g, and 7.16 (PPD), respectively. For the prediction of protein and moisture content in wheat flour, the R2, RMSE, and RPD values were 0.99, 0.29 mg/g, and 8.22 (protein), and 0.97, 0.22 mg/g, and 5.67 (moisture), respectively. The experimental results demonstrate that GA-OMTL outperforms three comparison methods in prediction accuracy, highlighting its potential for complex spectral tasks and confirming the practicality and robustness of the proposed model.
4087.
Jha, Abhash Kumar; Moradian, Shakiba; Krishnakumar, Arjun; Rapp, Martin; Hutter, Frank
$textbackslashtextttconfopt$: A Library for Implementation and Evaluation of Gradient-based One-Shot NAS Methods Proceedings Article
In: AutoML 2025 ABCD Track, 2025.
@inproceedings{<LineBreak>jha2025textttconfopt,
title = {$textbackslashtextttconfopt$: A Library for Implementation and Evaluation of Gradient-based One-Shot NAS Methods},
author = {Abhash Kumar Jha and Shakiba Moradian and Arjun Krishnakumar and Martin Rapp and Frank Hutter},
url = {https://openreview.net/forum?id=serEYBjyhK},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
booktitle = {AutoML 2025 ABCD Track},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
4086.
Zhao, Lingxiao; Zeng, Xianwen
BTSEG-Nas: a neural network architecture search-based multimodal MRI segmentation network for brain tumors Proceedings Article
In: Zhu, Peicheng; Lin, Guihua (Ed.): Fifth International Conference on Applied Mathematics, Modelling, and Intelligent Computing (CAMMIC 2025), pp. 136441X, International Society for Optics and Photonics SPIE, 2025.
@inproceedings{10.1117/12.3070322,
title = {BTSEG-Nas: a neural network architecture search-based multimodal MRI segmentation network for brain tumors},
author = {Lingxiao Zhao and Xianwen Zeng},
editor = {Peicheng Zhu and Guihua Lin},
url = {https://doi.org/10.1117/12.3070322},
doi = {10.1117/12.3070322},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
booktitle = {Fifth International Conference on Applied Mathematics, Modelling, and Intelligent Computing (CAMMIC 2025)},
volume = {13644},
pages = {136441X},
publisher = {SPIE},
organization = {International Society for Optics and Photonics},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
4085.
Huang, Junhao; Xue, Bing; Sun, Yanan; Zhang, Mengjie; Yen, Gary G.
LightMix: Multi-Objective Search for Lightweight Mixed-Scale Convolutional Neural Networks Journal Article
In: IEEE Transactions on Emerging Topics in Computational Intelligence, pp. 1-15, 2025.
@article{11023228,
title = {LightMix: Multi-Objective Search for Lightweight Mixed-Scale Convolutional Neural Networks},
author = {Junhao Huang and Bing Xue and Yanan Sun and Mengjie Zhang and Gary G. Yen},
url = {https://ieeexplore.ieee.org/abstract/document/11023228},
doi = {10.1109/TETCI.2025.3572041},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
journal = {IEEE Transactions on Emerging Topics in Computational Intelligence},
pages = {1-15},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
4084.
Hossain, Md Adnan Faisal; Zhu, Fengqing
Structured Pruning and Quantization for Learned Image Compression Technical Report
2025.
@techreport{hossain2025structuredpruningquantizationlearned,
title = {Structured Pruning and Quantization for Learned Image Compression},
author = {Md Adnan Faisal Hossain and Fengqing Zhu},
url = {https://arxiv.org/abs/2506.01229},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
4083.
Shen, Ruina; Peng, Zhangheng; Dong, Le; Wang, Jun; Dong, Weisheng; Shi, Guangming
Searching efficient network with lightweight optimization for image super-resolution Journal Article
In: Neurocomputing, vol. 647, pp. 130550, 2025, ISSN: 0925-2312.
@article{SHEN2025130550,
title = {Searching efficient network with lightweight optimization for image super-resolution},
author = {Ruina Shen and Zhangheng Peng and Le Dong and Jun Wang and Weisheng Dong and Guangming Shi},
url = {https://www.sciencedirect.com/science/article/pii/S0925231225012226},
doi = {https://doi.org/10.1016/j.neucom.2025.130550},
issn = {0925-2312},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
journal = {Neurocomputing},
volume = {647},
pages = {130550},
abstract = {Despite rapid advances in deep learning-based image super-resolution (SR), deploying them in resource-constrained scenarios (e.g., mobile devices) has remained a challenging task. Generally speaking, good performance of SR networks comes at the price of increased computational complexity and model parameters. To strike an improved trade-off, recent studies have considered the design of efficient SR networks. However, most existing methods only consider the network with fewer parameters and calculations, overlooking the actual inference latency time of the network on mobile devices. To meet this challenge of supporting real-time applications, a neural architecture search (NAS) method is developed to find an efficient SR network in this paper, where search space is designed, including efficient operations and lightweight attention blocks. Besides, a novel search method for channel numbers is proposed, called filter grouping search. And a subnet sampling-updating method is designed to accelerate the search. Finally, the weight pruning and adaptive weighted uncertainty-driven loss combined with contrastive distillation are used to optimize the final target network. The experimental results show that our method is superior to the existing state-of-the-art methods. Additionally, the proposed method is also tested on mobile devices and found that it has more advantages on the Kirin 990 5G ARM CPU, supporting its potential in mobile computing. The code is available at https://github.com/douzaikongcheng/EMNAS.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Despite rapid advances in deep learning-based image super-resolution (SR), deploying them in resource-constrained scenarios (e.g., mobile devices) has remained a challenging task. Generally speaking, good performance of SR networks comes at the price of increased computational complexity and model parameters. To strike an improved trade-off, recent studies have considered the design of efficient SR networks. However, most existing methods only consider the network with fewer parameters and calculations, overlooking the actual inference latency time of the network on mobile devices. To meet this challenge of supporting real-time applications, a neural architecture search (NAS) method is developed to find an efficient SR network in this paper, where search space is designed, including efficient operations and lightweight attention blocks. Besides, a novel search method for channel numbers is proposed, called filter grouping search. And a subnet sampling-updating method is designed to accelerate the search. Finally, the weight pruning and adaptive weighted uncertainty-driven loss combined with contrastive distillation are used to optimize the final target network. The experimental results show that our method is superior to the existing state-of-the-art methods. Additionally, the proposed method is also tested on mobile devices and found that it has more advantages on the Kirin 990 5G ARM CPU, supporting its potential in mobile computing. The code is available at https://github.com/douzaikongcheng/EMNAS.
4082.
Ji, Han; Feng, Yuqi; Fan, Jiahao; Sun, Yanan
CARL: Causality-guided Architecture Representation Learning for an Interpretable Performance Predictor Technical Report
2025.
@techreport{ji2025carlcausalityguidedarchitecturerepresentation,
title = {CARL: Causality-guided Architecture Representation Learning for an Interpretable Performance Predictor},
author = {Han Ji and Yuqi Feng and Jiahao Fan and Yanan Sun},
url = {https://arxiv.org/abs/2506.04001},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
4081.
Yu, Jiaqi; Gao, Yang; Yang, Hong; Tian, Zhihong; Zhang, Peng; Zhu, Xingquan
Automated graph anomaly detection with large language models Journal Article
In: Knowledge-Based Systems, vol. 324, pp. 113809, 2025, ISSN: 0950-7051.
@article{YU2025113809,
title = {Automated graph anomaly detection with large language models},
author = {Jiaqi Yu and Yang Gao and Hong Yang and Zhihong Tian and Peng Zhang and Xingquan Zhu},
url = {https://www.sciencedirect.com/science/article/pii/S095070512500855X},
doi = {https://doi.org/10.1016/j.knosys.2025.113809},
issn = {0950-7051},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
journal = {Knowledge-Based Systems},
volume = {324},
pages = {113809},
abstract = {Graph neural networks (GNNs) have emerged as powerful tools for graph anomaly detection (GAD). However, designing effective GNN architectures for GAD often demands considerable domain expertise and laborious manual tuning. Although graph neural architecture search (GNAS) has made significant progress in automating the discovery of effective deep architectures, existing GNAS methods are challenging to directly apply to GAD tasks due to the lack of a dedicated search space tailored for GAD and the difficulty in effectively incorporating domain expert knowledge into the model architecture generation process. To address these challenges, this paper proposes an automated graph anomaly detection (AutoGAD for short) framework. AutoGAD automates the generation of optimal neural network architectures through a predefined search space and an efficient search strategy. Specifically, we first design a novel search space tailored for GAD tasks based on the characteristics of the graph autoencoder framework. Then, we leverage a large language model (LLM) as the controller of GNAS, guiding the LLM to rapidly generate architectures suitable for GAD within the search space through well-designed prompts. Extensive experimental results demonstrate that AutoGAD can generate new architectures that outperform existing GAD models, and its effectiveness is consistently observed across different datasets.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Graph neural networks (GNNs) have emerged as powerful tools for graph anomaly detection (GAD). However, designing effective GNN architectures for GAD often demands considerable domain expertise and laborious manual tuning. Although graph neural architecture search (GNAS) has made significant progress in automating the discovery of effective deep architectures, existing GNAS methods are challenging to directly apply to GAD tasks due to the lack of a dedicated search space tailored for GAD and the difficulty in effectively incorporating domain expert knowledge into the model architecture generation process. To address these challenges, this paper proposes an automated graph anomaly detection (AutoGAD for short) framework. AutoGAD automates the generation of optimal neural network architectures through a predefined search space and an efficient search strategy. Specifically, we first design a novel search space tailored for GAD tasks based on the characteristics of the graph autoencoder framework. Then, we leverage a large language model (LLM) as the controller of GNAS, guiding the LLM to rapidly generate architectures suitable for GAD within the search space through well-designed prompts. Extensive experimental results demonstrate that AutoGAD can generate new architectures that outperform existing GAD models, and its effectiveness is consistently observed across different datasets.
4080.
Kumar, Swagat; Zaech, Jan-Nico; Wilmott, Colin Michael; Gool, Luc Van
RhoDARTS: Differentiable Quantum Architecture Search with Density Matrix Simulations Technical Report
2025.
@techreport{kumar2025rhodartsdifferentiablequantumarchitecture,
title = {RhoDARTS: Differentiable Quantum Architecture Search with Density Matrix Simulations},
author = {Swagat Kumar and Jan-Nico Zaech and Colin Michael Wilmott and Luc Van Gool},
url = {https://arxiv.org/abs/2506.03697},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
4079.
Coil, Collin; Cheney, Nick
What Makes Freezing Layers in Deep Neural Networks Effective? A Linear Separability Perspective Proceedings Article
In: AutoML 2025 Methods Track, 2025.
@inproceedings{<LineBreak>coil2025what,
title = {What Makes Freezing Layers in Deep Neural Networks Effective? A Linear Separability Perspective},
author = {Collin Coil and Nick Cheney},
url = {https://openreview.net/forum?id=DALK4KJTjX},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
booktitle = {AutoML 2025 Methods Track},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
4078.
Pava, Mateo Avila; Groh, René; Kist, Andreas M
EG-ENAS: Efficient and Generalizable Evolutionary Neural Architecture Search for Image Classification Proceedings Article
In: AutoML 2025 Methods Track, 2025.
@inproceedings{<LineBreak>pava2025egenas,
title = {EG-ENAS: Efficient and Generalizable Evolutionary Neural Architecture Search for Image Classification},
author = {Mateo Avila Pava and René Groh and Andreas M Kist},
url = {https://openreview.net/forum?id=3YWElIrU8a},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
booktitle = {AutoML 2025 Methods Track},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
4077.
Roshtkhari, Mehraveh Javan; Toews, Matthew; Pedersoli, Marco
Iterative Monte Carlo Tree Search for Neural Architecture Search Proceedings Article
In: AutoML 2025 Methods Track, 2025.
@inproceedings{<LineBreak>roshtkhari2025iterative,
title = {Iterative Monte Carlo Tree Search for Neural Architecture Search},
author = {Mehraveh Javan Roshtkhari and Matthew Toews and Marco Pedersoli},
url = {https://openreview.net/forum?id=GuwNztkceE},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
booktitle = {AutoML 2025 Methods Track},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
4076.
Chen, Xueqian; Mao, Zhaoyong; Chen, Zhiwei; Shen, Junge
EvoNAS4Battery: An Evolutionary NAS Approach for Predicting the Remaining Useful Life of Lithium-Ion Batteries Journal Article
In: IEEE Transactions on Instrumentation and Measurement, vol. 74, pp. 1-16, 2025.
@article{11021509,
title = {EvoNAS4Battery: An Evolutionary NAS Approach for Predicting the Remaining Useful Life of Lithium-Ion Batteries},
author = {Xueqian Chen and Zhaoyong Mao and Zhiwei Chen and Junge Shen},
url = {https://ieeexplore.ieee.org/abstract/document/11021509},
doi = {10.1109/TIM.2025.3573358},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
journal = {IEEE Transactions on Instrumentation and Measurement},
volume = {74},
pages = {1-16},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
4075.
Song, Wenhao; Wu, Xuan; Yang, Bo; Zhou, You; Xiao, Yubin; Liang, Yanchun; Ge, Hongwei; Lee, Heow Pueh; Wu, Chunguo
Towards Efficient Few-shot Graph Neural Architecture Search via Partitioning Gradient Contribution Technical Report
2025.
@techreport{song2025efficientfewshotgraphneural,
title = {Towards Efficient Few-shot Graph Neural Architecture Search via Partitioning Gradient Contribution},
author = {Wenhao Song and Xuan Wu and Bo Yang and You Zhou and Yubin Xiao and Yanchun Liang and Hongwei Ge and Heow Pueh Lee and Chunguo Wu},
url = {https://arxiv.org/abs/2506.01231},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
4074.
Zhou, Yuyang; Neri, Ferrante; Ong, Yew-Soon; Bai, Ruibin
SiamNAS: Siamese Surrogate Model for Dominance Relation Prediction in Multi-objective Neural Architecture Search Technical Report
2025.
@techreport{zhou2025siamnassiamesesurrogatemodel,
title = {SiamNAS: Siamese Surrogate Model for Dominance Relation Prediction in Multi-objective Neural Architecture Search},
author = {Yuyang Zhou and Ferrante Neri and Yew-Soon Ong and Ruibin Bai},
url = {https://arxiv.org/abs/2506.02623},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
4073.
Huang, Hongtao; Chang, Xiaojun; Yao, Lina
Flexiffusion: Training-Free Segment-Wise Neural Architecture Search for Efficient Diffusion Models Technical Report
2025.
@techreport{huang2025flexiffusiontrainingfreesegmentwiseneural,
title = {Flexiffusion: Training-Free Segment-Wise Neural Architecture Search for Efficient Diffusion Models},
author = {Hongtao Huang and Xiaojun Chang and Lina Yao},
url = {https://arxiv.org/abs/2506.02488},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
4072.
Garavagno, Andrea Mattia; Ragusa, Edoardo; Frisoli, Antonio; Gastaldo, Paolo
Searching Neural Architectures for Sensor Nodes on IoT Gateways Technical Report
2025.
@techreport{garavagno2025searchingneuralarchitecturessensor,
title = {Searching Neural Architectures for Sensor Nodes on IoT Gateways},
author = {Andrea Mattia Garavagno and Edoardo Ragusa and Antonio Frisoli and Paolo Gastaldo},
url = {https://arxiv.org/abs/2505.23939},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
4071.
Yu, Xinyi; Yin, Runan; Lin, Zhihao; Wang, Yongtao
ERF-NAS: Efficient Receptive Field-Based Zero-Shot NAS for Object Detection Proceedings Article
In: Bue, Alessio Del; Canton, Cristian; Pont-Tuset, Jordi; Tommasi, Tatiana (Ed.): Computer Vision – ECCV 2024 Workshops, pp. 219–234, Springer Nature Switzerland, Cham, 2025, ISBN: 978-3-031-91979-4.
@inproceedings{10.1007/978-3-031-91979-4_17,
title = {ERF-NAS: Efficient Receptive Field-Based Zero-Shot NAS for Object Detection},
author = {Xinyi Yu and Runan Yin and Zhihao Lin and Yongtao Wang},
editor = {Alessio Del Bue and Cristian Canton and Jordi Pont-Tuset and Tatiana Tommasi},
url = {https://link.springer.com/chapter/10.1007/978-3-031-91979-4_17},
isbn = {978-3-031-91979-4},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
booktitle = {Computer Vision – ECCV 2024 Workshops},
pages = {219–234},
publisher = {Springer Nature Switzerland},
address = {Cham},
abstract = {Neural Architecture Search (NAS) is widely used to design efficient and high-performance architectures for various models. However, most NAS methods for object detection tasks are complex and time-consuming, requiring substantial computational resources to train numerous candidate detectors in a vast search space. To address this issue, we propose ERF-NAS, an efficient receptive field-based zero-shot NAS method. First, we utilize the Effective Receptive Field (ERF) of the detector's backbone network as a zero-cost proxy to assess the candidate architecture's feature extraction quality and expression capabilities. The calculation of ERF only requires a single inference forward of a randomly initialized backbone network, eliminating the necessity for training. Second, to obtain better coordination between different network components, we introduce the transformation paradigm to adjust the depth of the neck network. To provide a more effective and accurate latency constraint of candidate architectures, we construct a latency table by measuring convolution operations in TensorRT format. Extensive experiments on MS COCO dataset show that ERF-NAS achieves superior accuracy-efficiency trade-off results with less than 1 GPU day search costs.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Neural Architecture Search (NAS) is widely used to design efficient and high-performance architectures for various models. However, most NAS methods for object detection tasks are complex and time-consuming, requiring substantial computational resources to train numerous candidate detectors in a vast search space. To address this issue, we propose ERF-NAS, an efficient receptive field-based zero-shot NAS method. First, we utilize the Effective Receptive Field (ERF) of the detector's backbone network as a zero-cost proxy to assess the candidate architecture's feature extraction quality and expression capabilities. The calculation of ERF only requires a single inference forward of a randomly initialized backbone network, eliminating the necessity for training. Second, to obtain better coordination between different network components, we introduce the transformation paradigm to adjust the depth of the neck network. To provide a more effective and accurate latency constraint of candidate architectures, we construct a latency table by measuring convolution operations in TensorRT format. Extensive experiments on MS COCO dataset show that ERF-NAS achieves superior accuracy-efficiency trade-off results with less than 1 GPU day search costs.
4070.
Hu, Mengxiang; Li, Junchi; Dong, Yongquan; Zhang, Zichen; Liu, Weifan; Zhang, Peilin; Ping, Yuchao; Jiang, Le; Yu, Zekuan
Mixed-GGNAS: Mixed Search-space NAS based on genetic algorithm combined with gradient descent for medical image segmentation Journal Article
In: Expert Systems with Applications, vol. 289, pp. 128338, 2025, ISSN: 0957-4174.
@article{HU2025128338,
title = {Mixed-GGNAS: Mixed Search-space NAS based on genetic algorithm combined with gradient descent for medical image segmentation},
author = {Mengxiang Hu and Junchi Li and Yongquan Dong and Zichen Zhang and Weifan Liu and Peilin Zhang and Yuchao Ping and Le Jiang and Zekuan Yu},
url = {https://www.sciencedirect.com/science/article/pii/S0957417425019578},
doi = {https://doi.org/10.1016/j.eswa.2025.128338},
issn = {0957-4174},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
journal = {Expert Systems with Applications},
volume = {289},
pages = {128338},
abstract = {Medical images segmentation is a pivotal procedure, playing a fundamental role in computer-assisted diagnosis and treatment. Despite the significant advancements in methods leveraging deep learning for this purpose, many networks still face challenges related to efficiency, often requiring substantial time and manual efforts. Neural architecture search (NAS) has gained considerable attention in the automated design of neural networks. This study introduces a new NAS method, Mixed-GGNAS, a Mixed Search-space NAS method based on Genetic algorithm combined with Gradient descent. Our approach creatively combines manually designed network blocks with DARTS blocks to construct a mixed search space. We then employ a method that integrates genetic algorithms and gradient descent to concurrently search for both block types and internal operations within the block. Within a U-shaped network framework, we propose a Multi-feature fusion strategy based on Vision Transformer (ViT) and search for hyperparameters of it. Additionally, we employ a Multi-scale mixed loss function to enhance the model’s ability to learn features at various scales. Experimental results demonstrate that the proposed approach outperforms or is comparable to the state-of-the-art NAS methods and manually designed Networks. Ablation studies conducted on two datasets further validate the method’s efficacy in enhancing model performance. The code is available at https://github.com/Hmxki/Mixed-GGNAS.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Medical images segmentation is a pivotal procedure, playing a fundamental role in computer-assisted diagnosis and treatment. Despite the significant advancements in methods leveraging deep learning for this purpose, many networks still face challenges related to efficiency, often requiring substantial time and manual efforts. Neural architecture search (NAS) has gained considerable attention in the automated design of neural networks. This study introduces a new NAS method, Mixed-GGNAS, a Mixed Search-space NAS method based on Genetic algorithm combined with Gradient descent. Our approach creatively combines manually designed network blocks with DARTS blocks to construct a mixed search space. We then employ a method that integrates genetic algorithms and gradient descent to concurrently search for both block types and internal operations within the block. Within a U-shaped network framework, we propose a Multi-feature fusion strategy based on Vision Transformer (ViT) and search for hyperparameters of it. Additionally, we employ a Multi-scale mixed loss function to enhance the model’s ability to learn features at various scales. Experimental results demonstrate that the proposed approach outperforms or is comparable to the state-of-the-art NAS methods and manually designed Networks. Ablation studies conducted on two datasets further validate the method’s efficacy in enhancing model performance. The code is available at https://github.com/Hmxki/Mixed-GGNAS.
4069.
Chen, Haibo; Wang, Xin; Zhang, Zeyang; Li, Haoyang; Feng, Ling; Zhu, Wenwu
AutoGFM: Automated Graph Foundation Model with Adaptive Architecture Customization Proceedings Article
In: Forty-second International Conference on Machine Learning, 2025.
@inproceedings{<LineBreak>chen2025autogfm,
title = {AutoGFM: Automated Graph Foundation Model with Adaptive Architecture Customization},
author = {Haibo Chen and Xin Wang and Zeyang Zhang and Haoyang Li and Ling Feng and Wenwu Zhu},
url = {https://openreview.net/forum?id=fCPB0qRJT2},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
booktitle = {Forty-second International Conference on Machine Learning},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
4068.
Amin, Md Hasibul; Mohammadi, Mohammadreza; Bakos, Jason D.; Zand, Ramtin
CrossNAS: A Cross-Layer Neural Architecture Search Framework for PIM Systems Technical Report
2025.
@techreport{amin2025crossnascrosslayerneuralarchitectureb,
title = {CrossNAS: A Cross-Layer Neural Architecture Search Framework for PIM Systems},
author = {Md Hasibul Amin and Mohammadreza Mohammadi and Jason D. Bakos and Ramtin Zand},
url = {https://arxiv.org/abs/2505.22868},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
4067.
WANG, Ke; SONG, Yafei; XU, Yunfei; QUAN, Wen; NI, Peng; WANG, Peng; LI, Chenghai; ZHI, Xinyan
A novel automated neural network architecture search method of air target intent recognition Journal Article
In: Chinese Journal of Aeronautics, vol. 38, no. 6, pp. 103295, 2025, ISSN: 1000-9361.
@article{WANG2025103295b,
title = {A novel automated neural network architecture search method of air target intent recognition},
author = {Ke WANG and Yafei SONG and Yunfei XU and Wen QUAN and Peng NI and Peng WANG and Chenghai LI and Xinyan ZHI},
url = {https://www.sciencedirect.com/science/article/pii/S1000936124004448},
doi = {https://doi.org/10.1016/j.cja.2024.11.005},
issn = {1000-9361},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
journal = {Chinese Journal of Aeronautics},
volume = {38},
number = {6},
pages = {103295},
abstract = {Modern air battlefield operations are characterized by flexibility and change, and the battlefield evolves rapidly and intricately. However, traditional air target intent recognition methods, which mainly rely on manually designed neural network models, find it difficult to maintain sustained and excellent performance in such a complex and changing environment. To address the problem of the adaptability of neural network models in complex environments, we propose a lightweight Transformer model (TransATIR) with a strong adaptive adjustment capability, based on the characteristics of air target intent recognition and the neural network architecture search technique. After conducting extensive experiments, it has been proved that TransATIR can efficiently extract the deep feature information from battlefield situation data by utilizing the neural architecture search algorithm, in order to quickly and accurately identify the real intention of the target. The experimental results indicate that TransATIR significantly improves recognition accuracy compared to the existing state-of-the-art methods, and also effectively reduces the computational complexity of the model.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Modern air battlefield operations are characterized by flexibility and change, and the battlefield evolves rapidly and intricately. However, traditional air target intent recognition methods, which mainly rely on manually designed neural network models, find it difficult to maintain sustained and excellent performance in such a complex and changing environment. To address the problem of the adaptability of neural network models in complex environments, we propose a lightweight Transformer model (TransATIR) with a strong adaptive adjustment capability, based on the characteristics of air target intent recognition and the neural network architecture search technique. After conducting extensive experiments, it has been proved that TransATIR can efficiently extract the deep feature information from battlefield situation data by utilizing the neural architecture search algorithm, in order to quickly and accurately identify the real intention of the target. The experimental results indicate that TransATIR significantly improves recognition accuracy compared to the existing state-of-the-art methods, and also effectively reduces the computational complexity of the model.
4066.
WANG, Ke; SONG, Yafei; XU, Yunfei; QUAN, Wen; NI, Peng; WANG, Peng; LI, Chenghai; ZHI, Xinyan
A novel automated neural network architecture search method of air target intent recognition Journal Article
In: Chinese Journal of Aeronautics, vol. 38, no. 6, pp. 103295, 2025, ISSN: 1000-9361.
@article{WANG2025103295,
title = {A novel automated neural network architecture search method of air target intent recognition},
author = {Ke WANG and Yafei SONG and Yunfei XU and Wen QUAN and Peng NI and Peng WANG and Chenghai LI and Xinyan ZHI},
url = {https://www.sciencedirect.com/science/article/pii/S1000936124004448},
doi = {https://doi.org/10.1016/j.cja.2024.11.005},
issn = {1000-9361},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
journal = {Chinese Journal of Aeronautics},
volume = {38},
number = {6},
pages = {103295},
abstract = {Modern air battlefield operations are characterized by flexibility and change, and the battlefield evolves rapidly and intricately. However, traditional air target intent recognition methods, which mainly rely on manually designed neural network models, find it difficult to maintain sustained and excellent performance in such a complex and changing environment. To address the problem of the adaptability of neural network models in complex environments, we propose a lightweight Transformer model (TransATIR) with a strong adaptive adjustment capability, based on the characteristics of air target intent recognition and the neural network architecture search technique. After conducting extensive experiments, it has been proved that TransATIR can efficiently extract the deep feature information from battlefield situation data by utilizing the neural architecture search algorithm, in order to quickly and accurately identify the real intention of the target. The experimental results indicate that TransATIR significantly improves recognition accuracy compared to the existing state-of-the-art methods, and also effectively reduces the computational complexity of the model.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Modern air battlefield operations are characterized by flexibility and change, and the battlefield evolves rapidly and intricately. However, traditional air target intent recognition methods, which mainly rely on manually designed neural network models, find it difficult to maintain sustained and excellent performance in such a complex and changing environment. To address the problem of the adaptability of neural network models in complex environments, we propose a lightweight Transformer model (TransATIR) with a strong adaptive adjustment capability, based on the characteristics of air target intent recognition and the neural network architecture search technique. After conducting extensive experiments, it has been proved that TransATIR can efficiently extract the deep feature information from battlefield situation data by utilizing the neural architecture search algorithm, in order to quickly and accurately identify the real intention of the target. The experimental results indicate that TransATIR significantly improves recognition accuracy compared to the existing state-of-the-art methods, and also effectively reduces the computational complexity of the model.
4065.
Amin, Md Hasibul; Mohammadi, Mohammadreza; Bakos, Jason D.; Zand, Ramtin
CrossNAS: A Cross-Layer Neural Architecture Search Framework for PIM Systems Technical Report
2025.
@techreport{amin2025crossnascrosslayerneuralarchitecture,
title = {CrossNAS: A Cross-Layer Neural Architecture Search Framework for PIM Systems},
author = {Md Hasibul Amin and Mohammadreza Mohammadi and Jason D. Bakos and Ramtin Zand},
url = {https://arxiv.org/abs/2505.22868},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
4064.
Xie, Yilin; Zhang, Shiqiang; Qing, Jixiang; Misener, Ruth; Tsay, Calvin
Global optimization of graph acquisition functions for neural architecture search Technical Report
2025.
@techreport{xie2025globaloptimizationgraphacquisition,
title = {Global optimization of graph acquisition functions for neural architecture search},
author = {Yilin Xie and Shiqiang Zhang and Jixiang Qing and Ruth Misener and Calvin Tsay},
url = {https://arxiv.org/abs/2505.23640},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
4063.
Sun, Yu; Zhang, Xianglin; Dong, Liang; Liu, Ning
In: Applied Soft Computing, vol. 179, pp. 113279, 2025, ISSN: 1568-4946.
@article{SUN2025113279,
title = {Multi-objective evolutionary neural architecture search for medical image analysis using transformer and large language models in advancing public health},
author = {Yu Sun and Xianglin Zhang and Liang Dong and Ning Liu},
url = {https://www.sciencedirect.com/science/article/pii/S1568494625005903},
doi = {https://doi.org/10.1016/j.asoc.2025.113279},
issn = {1568-4946},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
journal = {Applied Soft Computing},
volume = {179},
pages = {113279},
abstract = {The rapid growth of medical imaging data in modern healthcare networks demands sophisticated automated analysis methods that can maintain high accuracy while operating efficiently at scale. Current approaches using transformers and large language models (LLMs) face challenges balancing computational requirements with diagnostic precision across diverse healthcare settings. This paper presents TransMed-NAS (transformer medical neural architecture search), a multi-objective evolutionary neural architecture search framework that automatically discovers efficient hybrid architectures by integrating transformers and LLMs for medical image segmentation. Our approach leverages evolutionary computation to optimize segmentation accuracy and computational efficiency while incorporating medical domain knowledge through LLM guidance. The framework introduces several innovations: a hierarchical channel selection strategy that preserves clinically relevant features, a weight entanglement mechanism that accelerates architecture search through intelligent knowledge transfer, and a surrogate model acceleration technique that reduces computational overhead while maintaining reliability. Experimental results on the ISIC 2020 dataset demonstrate TransMed-NAS’s superior performance compared to state-of-the-art methods. Our small model variant achieves competitive accuracy (0.934 Dice score) with only 0.82M parameters, while our large variant establishes new benchmarks (0.947 Dice score) with significantly reduced computational requirements. Ablation studies confirm the effectiveness of each component, particularly highlighting how LLM integration enhances architecture search efficiency and clinical relevance. These results demonstrate TransMed-NAS’s potential to advance automated medical image analysis in resource-diverse healthcare settings, making sophisticated diagnostic capabilities more accessible to underserved communities.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
The rapid growth of medical imaging data in modern healthcare networks demands sophisticated automated analysis methods that can maintain high accuracy while operating efficiently at scale. Current approaches using transformers and large language models (LLMs) face challenges balancing computational requirements with diagnostic precision across diverse healthcare settings. This paper presents TransMed-NAS (transformer medical neural architecture search), a multi-objective evolutionary neural architecture search framework that automatically discovers efficient hybrid architectures by integrating transformers and LLMs for medical image segmentation. Our approach leverages evolutionary computation to optimize segmentation accuracy and computational efficiency while incorporating medical domain knowledge through LLM guidance. The framework introduces several innovations: a hierarchical channel selection strategy that preserves clinically relevant features, a weight entanglement mechanism that accelerates architecture search through intelligent knowledge transfer, and a surrogate model acceleration technique that reduces computational overhead while maintaining reliability. Experimental results on the ISIC 2020 dataset demonstrate TransMed-NAS’s superior performance compared to state-of-the-art methods. Our small model variant achieves competitive accuracy (0.934 Dice score) with only 0.82M parameters, while our large variant establishes new benchmarks (0.947 Dice score) with significantly reduced computational requirements. Ablation studies confirm the effectiveness of each component, particularly highlighting how LLM integration enhances architecture search efficiency and clinical relevance. These results demonstrate TransMed-NAS’s potential to advance automated medical image analysis in resource-diverse healthcare settings, making sophisticated diagnostic capabilities more accessible to underserved communities.
4062.
Wang, Chen; Guo, Tiezheng; Yang, Qingwen; Liu, Yanyi; Tang, Jiawei; Wen, Yingyou
A NAS-Based Risk Prediction Model and Interpretable System for Amyloidosis Journal Article
In: Computers, Materials and Continua, vol. 83, no. 3, pp. 5561-5574, 2025, ISSN: 1546-2218.
@article{WANG20255561,
title = {A NAS-Based Risk Prediction Model and Interpretable System for Amyloidosis},
author = {Chen Wang and Tiezheng Guo and Qingwen Yang and Yanyi Liu and Jiawei Tang and Yingyou Wen},
url = {https://www.sciencedirect.com/science/article/pii/S1546221825004837},
doi = {https://doi.org/10.32604/cmc.2025.063676},
issn = {1546-2218},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
journal = {Computers, Materials and Continua},
volume = {83},
number = {3},
pages = {5561-5574},
abstract = {Primary light chain amyloidosis is a rare hematologic disease with multi-organ involvement. Nearly one-third of patients with amyloidosis experience five or more consultations before diagnosis, which may lead to a poor prognosis due to delayed diagnosis. Early risk prediction based on artificial intelligence is valuable for clinical diagnosis and treatment of amyloidosis. For this disease, we propose an Evolutionary Neural Architecture Searching (ENAS) based risk prediction model, which achieves high-precision early risk prediction using physical examination data as a reference factor. To further enhance the value of clinic application, we designed a natural language-based interpretable system around the NAS-assisted risk prediction model for amyloidosis, which utilizes a large language model and Retrieval-Augmented Generation (RAG) to achieve further interpretation of the predicted conclusions. We also propose a document-based global semantic slicing approach in RAG to achieve more accurate slicing and improve the professionalism of the generated interpretations. Tests and implementation show that the proposed risk prediction model can be effectively used for early screening of amyloidosis and that the interpretation method based on the large language model and RAG can effectively provide professional interpretation of predicted results, which provides an effective method and means for the clinical applications of AI.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Primary light chain amyloidosis is a rare hematologic disease with multi-organ involvement. Nearly one-third of patients with amyloidosis experience five or more consultations before diagnosis, which may lead to a poor prognosis due to delayed diagnosis. Early risk prediction based on artificial intelligence is valuable for clinical diagnosis and treatment of amyloidosis. For this disease, we propose an Evolutionary Neural Architecture Searching (ENAS) based risk prediction model, which achieves high-precision early risk prediction using physical examination data as a reference factor. To further enhance the value of clinic application, we designed a natural language-based interpretable system around the NAS-assisted risk prediction model for amyloidosis, which utilizes a large language model and Retrieval-Augmented Generation (RAG) to achieve further interpretation of the predicted conclusions. We also propose a document-based global semantic slicing approach in RAG to achieve more accurate slicing and improve the professionalism of the generated interpretations. Tests and implementation show that the proposed risk prediction model can be effectively used for early screening of amyloidosis and that the interpretation method based on the large language model and RAG can effectively provide professional interpretation of predicted results, which provides an effective method and means for the clinical applications of AI.
4061.
Wei, Lan; Zhang, Dandan
A Dataset and Benchmarks for Deep Learning-Based Optical Microrobot Pose and Depth Perception Technical Report
2025.
@techreport{wei2025datasetbenchmarksdeeplearningbased,
title = {A Dataset and Benchmarks for Deep Learning-Based Optical Microrobot Pose and Depth Perception},
author = {Lan Wei and Dandan Zhang},
url = {https://arxiv.org/abs/2505.18303},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
4060.
Aach, Marcel; Blanc, Cyril; Lintermann, Andreas; Grave, Kurt De
Optimizing edge AI models on HPC systems with the edge in the loop Technical Report
2025.
@techreport{aach2025optimizingedgeaimodels,
title = {Optimizing edge AI models on HPC systems with the edge in the loop},
author = {Marcel Aach and Cyril Blanc and Andreas Lintermann and Kurt De Grave},
url = {https://arxiv.org/abs/2505.19995},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
4059.
Xie, Qing; Yu, Ruiyun
Expo-GAN: A Style Transfer Generative Adversarial Network for Exhibition Hall Design Based on Optimized Cyclic and Neural Architecture Search Journal Article
In: Computers, Materials and Continua, vol. 83, no. 3, pp. 4757-4774, 2025, ISSN: 1546-2218.
@article{XIE20254757,
title = {Expo-GAN: A Style Transfer Generative Adversarial Network for Exhibition Hall Design Based on Optimized Cyclic and Neural Architecture Search},
author = {Qing Xie and Ruiyun Yu},
url = {https://www.sciencedirect.com/science/article/pii/S1546221825004138},
doi = {https://doi.org/10.32604/cmc.2025.063345},
issn = {1546-2218},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
journal = {Computers, Materials and Continua},
volume = {83},
number = {3},
pages = {4757-4774},
abstract = {This study presents a groundbreaking method named Expo-GAN (Exposition-Generative Adversarial Network) for style transfer in exhibition hall design, using a refined version of the Cycle Generative Adversarial Network (CycleGAN). The primary goal is to enhance the transformation of image styles while maintaining visual consistency, an area where current CycleGAN models often fall short. These traditional models typically face difficulties in accurately capturing expansive features as well as the intricate stylistic details necessary for high-quality image transformation. To address these limitations, the research introduces several key modifications to the CycleGAN architecture. Enhancements to the generator involve integrating U-net with SpecTransformer modules. This integration incorporates the use of Fourier transform techniques coupled with multi-head self-attention mechanisms, which collectively improve the generator’s ability to depict both large-scale structural patterns and minute elements meticulously in the generated images. This enhancement allows the generator to achieve a more detailed and coherent fusion of styles, essential for exhibition hall designs where both broad aesthetic strokes and detailed nuances matter significantly. The study also proposes innovative changes to the discriminator by employing dilated convolution and global attention mechanisms. These are derived using the Differentiable Architecture Search (DARTS) Neural Architecture Search framework to expand the receptive field, which is crucial for recognizing comprehensive artistically styled images. By broadening the ability to discern complex artistic features, the model avoids previous pitfalls associated with style inconsistency and missing detailed features. Moreover, the traditional cyde-consistency loss function is replaced with the Learned Perceptual Image Patch Similarity (LPIPS) metric. This shift aims to significantly enhance the perceptual quality of the resultant images by prioritizing human-perceived similarities, which aligns better with user expectations and professional standards in design aesthetics. The experimental phase of this research demonstrates that this novel approach consistently outperforms the conventional CycleGAN across a broad range of datasets. Complementary ablation studies and qualitative assessments underscore its superiority, particularly in maintaining detail fidelity and style continuity. This is critical for creating a visually harmonious exhibition hall design where every detail contributes to the overall aesthetic appeal. The results illustrate that this refined approach effectively bridges the gap between technical capability and artistic necessity, marking a significant advancement in computational design methodologies.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
This study presents a groundbreaking method named Expo-GAN (Exposition-Generative Adversarial Network) for style transfer in exhibition hall design, using a refined version of the Cycle Generative Adversarial Network (CycleGAN). The primary goal is to enhance the transformation of image styles while maintaining visual consistency, an area where current CycleGAN models often fall short. These traditional models typically face difficulties in accurately capturing expansive features as well as the intricate stylistic details necessary for high-quality image transformation. To address these limitations, the research introduces several key modifications to the CycleGAN architecture. Enhancements to the generator involve integrating U-net with SpecTransformer modules. This integration incorporates the use of Fourier transform techniques coupled with multi-head self-attention mechanisms, which collectively improve the generator’s ability to depict both large-scale structural patterns and minute elements meticulously in the generated images. This enhancement allows the generator to achieve a more detailed and coherent fusion of styles, essential for exhibition hall designs where both broad aesthetic strokes and detailed nuances matter significantly. The study also proposes innovative changes to the discriminator by employing dilated convolution and global attention mechanisms. These are derived using the Differentiable Architecture Search (DARTS) Neural Architecture Search framework to expand the receptive field, which is crucial for recognizing comprehensive artistically styled images. By broadening the ability to discern complex artistic features, the model avoids previous pitfalls associated with style inconsistency and missing detailed features. Moreover, the traditional cyde-consistency loss function is replaced with the Learned Perceptual Image Patch Similarity (LPIPS) metric. This shift aims to significantly enhance the perceptual quality of the resultant images by prioritizing human-perceived similarities, which aligns better with user expectations and professional standards in design aesthetics. The experimental phase of this research demonstrates that this novel approach consistently outperforms the conventional CycleGAN across a broad range of datasets. Complementary ablation studies and qualitative assessments underscore its superiority, particularly in maintaining detail fidelity and style continuity. This is critical for creating a visually harmonious exhibition hall design where every detail contributes to the overall aesthetic appeal. The results illustrate that this refined approach effectively bridges the gap between technical capability and artistic necessity, marking a significant advancement in computational design methodologies.
4058.
Zechen, Zheng; Xuelei, He; Fengjun, Zhao; Xiaowei, He
PSNAS-Net: Hybrid gradient-physical optimizationfor efficient neural architecture search in customized medical imaging analysis Journal Article
In: Expert Systems with Applications, vol. 288, pp. 128155, 2025, ISSN: 0957-4174.
@article{ZECHEN2025128155,
title = {PSNAS-Net: Hybrid gradient-physical optimizationfor efficient neural architecture search in customized medical imaging analysis},
author = {Zheng Zechen and He Xuelei and Zhao Fengjun and He Xiaowei},
url = {https://www.sciencedirect.com/science/article/pii/S0957417425017750},
doi = {https://doi.org/10.1016/j.eswa.2025.128155},
issn = {0957-4174},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
journal = {Expert Systems with Applications},
volume = {288},
pages = {128155},
abstract = {Neural architecture search (NAS) facilitates the automated construction of neural networks tailored to specific tasks and requirements, resulting in models that are more closely aligned with the target task’s demands. However, in many studies, the extensive design space, high search costs, and time-consuming evaluation calculations render NAS impractical for numerous medical data tasks. Addressing these challenges, this study introduces an efficient algorithm for searching deep learning architectures. Initially, we propose 19 fundamental rules to streamline the design space, thereby reducing its scale. To improve the efficiency of the algorithm, we designed a NAS framework (PSNAS-Net) for convolutional neural networks and VisionTransformer, which consists of two search stages: Firstly, the improved powell algorithm is used to determine the model range, and the population-based simulated annealing algorithm is utilized to expedite the search for the final model. During the neural architecture search process, we consider accuracy, parameters, FLOPs, and model stability as comprehensive evaluation objectives, we designed a robust, flexible, and comprehensive metric for model evaluation. The experimental results demonstrate that PSNAS-Net achieves significantly shorter search times (0.05-1.47 GPU Days) compared to 19 existing NAS methods, while discovering compact models (as small as 0.11M) with superior performance across five medical image benchmarks. This study offers a viable approach for model search that accommodates individualized requirements.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Neural architecture search (NAS) facilitates the automated construction of neural networks tailored to specific tasks and requirements, resulting in models that are more closely aligned with the target task’s demands. However, in many studies, the extensive design space, high search costs, and time-consuming evaluation calculations render NAS impractical for numerous medical data tasks. Addressing these challenges, this study introduces an efficient algorithm for searching deep learning architectures. Initially, we propose 19 fundamental rules to streamline the design space, thereby reducing its scale. To improve the efficiency of the algorithm, we designed a NAS framework (PSNAS-Net) for convolutional neural networks and VisionTransformer, which consists of two search stages: Firstly, the improved powell algorithm is used to determine the model range, and the population-based simulated annealing algorithm is utilized to expedite the search for the final model. During the neural architecture search process, we consider accuracy, parameters, FLOPs, and model stability as comprehensive evaluation objectives, we designed a robust, flexible, and comprehensive metric for model evaluation. The experimental results demonstrate that PSNAS-Net achieves significantly shorter search times (0.05-1.47 GPU Days) compared to 19 existing NAS methods, while discovering compact models (as small as 0.11M) with superior performance across five medical image benchmarks. This study offers a viable approach for model search that accommodates individualized requirements.
4057.
Becktepe, Jannis; Hennig, Leona; Oeltze-Jafra, Steffen; Lindauer, Marius
Auto-nnU-Net: Towards Automated Medical Image Segmentation Technical Report
2025.
@techreport{becktepe2025autonnunetautomatedmedicalimage,
title = {Auto-nnU-Net: Towards Automated Medical Image Segmentation},
author = {Jannis Becktepe and Leona Hennig and Steffen Oeltze-Jafra and Marius Lindauer},
url = {https://arxiv.org/abs/2505.16561},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
4056.
Pinos, Michal; Klhufek, Jan; Mrazek, Vojtech; Sekanina, Lukas
Inference Energy Analysis in Context of Hardware-Aware NAS Proceedings Article
In: 2025 IEEE 28th International Symposium on Design and Diagnostics of Electronic Circuits and Systems (DDECS), pp. 161-164, 2025.
@inproceedings{11006674,
title = {Inference Energy Analysis in Context of Hardware-Aware NAS},
author = {Michal Pinos and Jan Klhufek and Vojtech Mrazek and Lukas Sekanina},
url = {https://ieeexplore.ieee.org/abstract/document/11006674},
doi = {10.1109/DDECS63720.2025.11006674},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
booktitle = {2025 IEEE 28th International Symposium on Design and Diagnostics of Electronic Circuits and Systems (DDECS)},
pages = {161-164},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
4055.
Louis, Gani; Caves, Kenett; Bello, Sadis
Hybrid Neural Network Architectures: Integrating Convolutional and Recurrent Layers with Genetic Optimization for Anomaly Detection Journal Article
In: 2025.
@article{articlem,
title = {Hybrid Neural Network Architectures: Integrating Convolutional and Recurrent Layers with Genetic Optimization for Anomaly Detection},
author = {Gani Louis and Kenett Caves and Sadis Bello},
url = {https://www.researchgate.net/profile/Sadis-Bello/publication/391950569_Hybrid_Neural_Network_Architectures_Integrating_Convolutional_and_Recurrent_Layers_with_Genetic_Optimization_for_Anomaly_Detection/links/682e59a78a76251f22e4adfb/Hybrid-Neural-Network-Architectures-Integrating-Convolutional-and-Recurrent-Layers-with-Genetic-Optimization-for-Anomaly-Detection.pdf},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
4054.
Phan, Quan Minh; Luong, Ngoc Hoang
From Hand-Crafted Metrics to Evolved Training-Free Performance Predictors for Neural Architecture Search via Genetic Programming Technical Report
2025.
@techreport{phan2025handcraftedmetricsevolvedtrainingfree,
title = {From Hand-Crafted Metrics to Evolved Training-Free Performance Predictors for Neural Architecture Search via Genetic Programming},
author = {Quan Minh Phan and Ngoc Hoang Luong},
url = {https://arxiv.org/abs/2505.15832},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
4053.
Zheng, Jie; He, Chunlin; Man, Wenxing; Wang, Jing
Training-free multi-scale neural architecture search for high-incidence cancer prediction Journal Article
In: Engineering Applications of Artificial Intelligence, vol. 156, pp. 111089, 2025, ISSN: 0952-1976.
@article{ZHENG2025111089,
title = {Training-free multi-scale neural architecture search for high-incidence cancer prediction},
author = {Jie Zheng and Chunlin He and Wenxing Man and Jing Wang},
url = {https://www.sciencedirect.com/science/article/pii/S0952197625010905},
doi = {https://doi.org/10.1016/j.engappai.2025.111089},
issn = {0952-1976},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
journal = {Engineering Applications of Artificial Intelligence},
volume = {156},
pages = {111089},
abstract = {Deep neural networks excel in high-incidence cancer prediction; however, designing networks that predict specific cancers is time-consuming and requires expert. The neural architecture search method offers a way to automate network design and has shown success in natural image. However, the small and varying lesion sizes in cancer image pose challenges, and most neural architecture search methods are computationally expensive and exhibit low agent correlation. Therefore, we propose a training-free multi-scale neural architecture search method for high-incidence cancer prediction. We introduce a multi-scale search space to address varying lesion sizes; and identify optimal scale combinations for feature extraction. To reduce computational costs and improve agent correlation, we design a training-free agent that evaluates network performance based on convergence, expressiveness, trainability, and complexity, enabling efficient neural architecture search implementation. Our extensive experiments on the NAS-Bench-201, MedmnistV2, LC25000, BreakHis, and CRC-5000 datasets show that our method outperforms both manually designed networks and state-of-the-art neural architecture search methods. The results demonstrate average improvements of 4.2%, 1.88%, 79.45%, 34.31%, and 31.71% in accuracy, area under the curve, search time, and Kendall and Spearman correlation coefficients, respectively.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Deep neural networks excel in high-incidence cancer prediction; however, designing networks that predict specific cancers is time-consuming and requires expert. The neural architecture search method offers a way to automate network design and has shown success in natural image. However, the small and varying lesion sizes in cancer image pose challenges, and most neural architecture search methods are computationally expensive and exhibit low agent correlation. Therefore, we propose a training-free multi-scale neural architecture search method for high-incidence cancer prediction. We introduce a multi-scale search space to address varying lesion sizes; and identify optimal scale combinations for feature extraction. To reduce computational costs and improve agent correlation, we design a training-free agent that evaluates network performance based on convergence, expressiveness, trainability, and complexity, enabling efficient neural architecture search implementation. Our extensive experiments on the NAS-Bench-201, MedmnistV2, LC25000, BreakHis, and CRC-5000 datasets show that our method outperforms both manually designed networks and state-of-the-art neural architecture search methods. The results demonstrate average improvements of 4.2%, 1.88%, 79.45%, 34.31%, and 31.71% in accuracy, area under the curve, search time, and Kendall and Spearman correlation coefficients, respectively.
4052.
Kocher, Nick; Wassermann, Christian; Hennig, Leona; Seng, Jonas; Hoos, Holger; Kersting, Kristian; Lindauer, Marius; Müller, Matthias
Guidelines for the Quality Assessment of Energy-Aware NAS Benchmarks Technical Report
2025.
@techreport{kocher2025guidelinesqualityassessmentenergyaware,
title = {Guidelines for the Quality Assessment of Energy-Aware NAS Benchmarks},
author = {Nick Kocher and Christian Wassermann and Leona Hennig and Jonas Seng and Holger Hoos and Kristian Kersting and Marius Lindauer and Matthias Müller},
url = {https://arxiv.org/abs/2505.15631},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
4051.
Ardila, Diego Páez; Carvalho, Thiago; Saavedra, Santiago Vasquez; Niño, Cesar Valencia; Figueiredo, Karla; Vellasco, Marley
Quantum-Inspired NAS With Attention-Based Search Spaces in Medical Applications Proceedings Article
In: 2025 IEEE Symposium on Computational Intelligence in Health and Medicine Companion (CIHM Companion), pp. 1-5, 2025.
@inproceedings{11002695,
title = {Quantum-Inspired NAS With Attention-Based Search Spaces in Medical Applications},
author = {Diego Páez Ardila and Thiago Carvalho and Santiago Vasquez Saavedra and Cesar Valencia Niño and Karla Figueiredo and Marley Vellasco},
doi = {10.1109/CIHMCompanion65205.2025.11002695},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
booktitle = {2025 IEEE Symposium on Computational Intelligence in Health and Medicine Companion (CIHM Companion)},
pages = {1-5},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
4050.
Cao, Bin; Deng, Huanyu; Hao, Yiming; Luo, Xiao
Multi-view information fusion based on federated multi-objective neural architecture search for MRI semantic segmentation Journal Article
In: Information Fusion, vol. 123, pp. 103301, 2025, ISSN: 1566-2535.
@article{CAO2025103301,
title = {Multi-view information fusion based on federated multi-objective neural architecture search for MRI semantic segmentation},
author = {Bin Cao and Huanyu Deng and Yiming Hao and Xiao Luo},
url = {https://www.sciencedirect.com/science/article/pii/S1566253525003744},
doi = {https://doi.org/10.1016/j.inffus.2025.103301},
issn = {1566-2535},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
journal = {Information Fusion},
volume = {123},
pages = {103301},
abstract = {With the rapid development of artificial intelligence, medical image semantic segmentation is being used more widely. However, centralized training can lead to privacy risks. At the same time, MRI provides multiple views that together describe the anatomical structure of a lesion, but a single view may not fully capture all features. Therefore, integrating multi-view information in a federated learning setting is a key challenge for improving model generalization. This study combines federated learning, neural architecture search (NAS) and data fusion techniques to address issues related to data privacy, cross-institutional data distribution differences and multi-view information fusion in medical imaging. To achieve this, we propose the FL-MONAS framework, which leverages the advantages of NAS and federated learning. It uses a Pareto-frontier-based multi-objective optimization strategy to effectively combine 2D MRI with 3D anatomical structures, improving model performance while ensuring data privacy. Experimental results show that FL-MONAS maintains strong segmentation performance even in non-IID scenarios, providing an efficient and privacy-friendly solution for medical image analysis.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
With the rapid development of artificial intelligence, medical image semantic segmentation is being used more widely. However, centralized training can lead to privacy risks. At the same time, MRI provides multiple views that together describe the anatomical structure of a lesion, but a single view may not fully capture all features. Therefore, integrating multi-view information in a federated learning setting is a key challenge for improving model generalization. This study combines federated learning, neural architecture search (NAS) and data fusion techniques to address issues related to data privacy, cross-institutional data distribution differences and multi-view information fusion in medical imaging. To achieve this, we propose the FL-MONAS framework, which leverages the advantages of NAS and federated learning. It uses a Pareto-frontier-based multi-objective optimization strategy to effectively combine 2D MRI with 3D anatomical structures, improving model performance while ensuring data privacy. Experimental results show that FL-MONAS maintains strong segmentation performance even in non-IID scenarios, providing an efficient and privacy-friendly solution for medical image analysis.
4049.
Chabal, Daphnee; Muller, Tim; Zhang, Eloise; Sapra, Dolly; Laat, Cees; Mann, Zoltán Ádám
COLIBRI: Optimizing Multi-party Secure Neural Network Inference Time for Transformers Proceedings Article
In: Zlatolas, Lili Nemec; Rannenberg, Kai; Welzer, Tatjana; Garcia-Alfaro, Joaquin (Ed.): ICT Systems Security and Privacy Protection, pp. 17–31, Springer Nature Switzerland, Cham, 2025, ISBN: 978-3-031-92882-6.
@inproceedings{10.1007/978-3-031-92882-6_2,
title = {COLIBRI: Optimizing Multi-party Secure Neural Network Inference Time for Transformers},
author = {Daphnee Chabal and Tim Muller and Eloise Zhang and Dolly Sapra and Cees Laat and Zoltán Ádám Mann},
editor = {Lili Nemec Zlatolas and Kai Rannenberg and Tatjana Welzer and Joaquin Garcia-Alfaro},
isbn = {978-3-031-92882-6},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
booktitle = {ICT Systems Security and Privacy Protection},
pages = {17–31},
publisher = {Springer Nature Switzerland},
address = {Cham},
abstract = {Secure Neural Network Inference (SNNI) protocols enable privacy-preserving inference by ensuring the confidentiality of inputs, model weights, and outputs. However, large neural networks, particularly Transformers, face significant challenges in SNNI due to high computational costs and slow execution, as these networks are typically optimized for accuracy rather than secure inference speed. We present COLIBRI, a novel approach that optimizes neural networks for efficient SNNI using Neural Architecture Search (NAS). Unlike prior methods, COLIBRI directly incorporates SNNI execution time as an optimization objective, leveraging a prediction model to estimate execution time without repeatedly running costly SNNI protocols during NAS. Our results on Cityscapes, a complex image segmentation task, show that COLIBRI reduces SNNI execution time by 26–33% while maintaining accuracy, marking a significant advancement in secure AI deployment.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Secure Neural Network Inference (SNNI) protocols enable privacy-preserving inference by ensuring the confidentiality of inputs, model weights, and outputs. However, large neural networks, particularly Transformers, face significant challenges in SNNI due to high computational costs and slow execution, as these networks are typically optimized for accuracy rather than secure inference speed. We present COLIBRI, a novel approach that optimizes neural networks for efficient SNNI using Neural Architecture Search (NAS). Unlike prior methods, COLIBRI directly incorporates SNNI execution time as an optimization objective, leveraging a prediction model to estimate execution time without repeatedly running costly SNNI protocols during NAS. Our results on Cityscapes, a complex image segmentation task, show that COLIBRI reduces SNNI execution time by 26–33% while maintaining accuracy, marking a significant advancement in secure AI deployment.
4048.
Rumiantsev, Pavel; Coates, Mark
Half Search Space is All You Need Technical Report
2025.
@techreport{rumiantsev2025halfsearchspaceneed,
title = {Half Search Space is All You Need},
author = {Pavel Rumiantsev and Mark Coates},
url = {https://arxiv.org/abs/2505.13586},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
4047.
Trirat, Patara; Lee, Jae-Gil
MONAQ: Multi-Objective Neural Architecture Querying for Time-Series Analysis on Resource-Constrained Devices Technical Report
2025.
@techreport{trirat2025monaqmultiobjectiveneuralarchitecture,
title = {MONAQ: Multi-Objective Neural Architecture Querying for Time-Series Analysis on Resource-Constrained Devices},
author = {Patara Trirat and Jae-Gil Lee},
url = {https://arxiv.org/abs/2505.10607},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
4046.
Hoang, Trong-Minh; Pham, Tuan-Anh; van-Nhan Nguyen,; Doan, Duc-Thang; Dao, Nhu-Ngoc
Leveraging Edge Intelligence for Solar Energy Management in Smart Grids Journal Article
In: IEEE Access, vol. 13, pp. 88093-88104, 2025.
@article{11005531,
title = {Leveraging Edge Intelligence for Solar Energy Management in Smart Grids},
author = {Trong-Minh Hoang and Tuan-Anh Pham and van-Nhan Nguyen and Duc-Thang Doan and Nhu-Ngoc Dao},
url = {https://ieeexplore.ieee.org/document/11005531},
doi = {10.1109/ACCESS.2025.3570595},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
journal = {IEEE Access},
volume = {13},
pages = {88093-88104},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
4045.
Gambella, Matteo; Solar, Vicente Javier Castro; Roveri, Manuel
SEAL: Searching Expandable Architectures for Incremental Learning Technical Report
2025.
@techreport{gambella2025sealsearchingexpandablearchitectures,
title = {SEAL: Searching Expandable Architectures for Incremental Learning},
author = {Matteo Gambella and Vicente Javier Castro Solar and Manuel Roveri},
url = {https://arxiv.org/abs/2505.10457},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
4044.
Chen, Samuel Yen-Chi; Liu, Chen-Yu; Chen, Kuan-Cheng; Huang, Wei-Jia; Chang, Yen-Jui; Huang, Wei-Hao
Differentiable Quantum Architecture Search in Quantum-Enhanced Neural Network Parameter Generation Technical Report
2025.
@techreport{chen2025differentiablequantumarchitecturesearch,
title = {Differentiable Quantum Architecture Search in Quantum-Enhanced Neural Network Parameter Generation},
author = {Samuel Yen-Chi Chen and Chen-Yu Liu and Kuan-Cheng Chen and Wei-Jia Huang and Yen-Jui Chang and Wei-Hao Huang},
url = {https://arxiv.org/abs/2505.09653},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
4043.
Wang, Yancheng; Jojic, Nebojsa; Yang, Yingzhen
Differentiable Channel Selection in Self-Attention For Person Re-Identification Technical Report
2025.
@techreport{wang2025differentiablechannelselectionselfattention,
title = {Differentiable Channel Selection in Self-Attention For Person Re-Identification},
author = {Yancheng Wang and Nebojsa Jojic and Yingzhen Yang},
url = {https://arxiv.org/abs/2505.08961},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
4042.
Deepa, S.; S, Parthiban; S, Angel; M, Divyalakshmi
Deep Analysis and Detection of Skin Disease using YOLO-NAS Algorithm Proceedings Article
In: 2025 5th International Conference on Trends in Material Science and Inventive Materials (ICTMIM), pp. 1626-1631, 2025.
@inproceedings{10987944,
title = {Deep Analysis and Detection of Skin Disease using YOLO-NAS Algorithm},
author = {S. Deepa and Parthiban S and Angel S and Divyalakshmi M},
url = {https://ieeexplore.ieee.org/abstract/document/10987944},
doi = {10.1109/ICTMIM65579.2025.10987944},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
booktitle = {2025 5th International Conference on Trends in Material Science and Inventive Materials (ICTMIM)},
pages = {1626-1631},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
4041.
Mecharbat, Lotfi Abdelkrim; Almakky, Ibrahim; Takac, Martin; Yaqub, Mohammad
MedNNS: Supernet-based Medical Task-Adaptive Neural Network Search Technical Report
2025.
@techreport{mecharbat2025mednnssupernetbasedmedicaltaskadaptiveb,
title = {MedNNS: Supernet-based Medical Task-Adaptive Neural Network Search},
author = {Lotfi Abdelkrim Mecharbat and Ibrahim Almakky and Martin Takac and Mohammad Yaqub},
url = {https://arxiv.org/abs/2504.15865},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
4040.
Poddenige, Deshani Geethika; Seneviratne, Sachith; Senanayake, Damith; Niranjan, Mahesan; Suganthan, PN; Halgamuge, Saman
Arch-LLM: Taming LLMs for Neural Architecture Generation via Unsupervised Discrete Representation Learning Technical Report
2025.
@techreport{poddenige2025archllmtamingllmsneuralb,
title = {Arch-LLM: Taming LLMs for Neural Architecture Generation via Unsupervised Discrete Representation Learning},
author = {Deshani Geethika Poddenige and Sachith Seneviratne and Damith Senanayake and Mahesan Niranjan and PN Suganthan and Saman Halgamuge},
url = {https://arxiv.org/abs/2503.22063},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}