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.
0000
38.
Loni, Mohammad; Mousavi, Hamid; Riazati, Mohammad; Daneshtalab, Masoud; Sjödin, Mikael
TAS : Ternarized Neural Architecture Search for Resource-Constrained Edge Devices Proceedings Article
In: Design, Automation and Test in Europe ConferenceDesign, Automation and Test in Europe Conference (DATE) 2022, ANTWERP, BELGIUM :, 0000.
@inproceedings{Loni1620831,
title = {TAS : Ternarized Neural Architecture Search for Resource-Constrained Edge Devices},
author = {Mohammad Loni and Hamid Mousavi and Mohammad Riazati and Masoud Daneshtalab and Mikael Sjödin},
url = {https://www.diva-portal.org/smash/record.jsf?pid=diva2%3A1620831&dswid=-1720},
booktitle = {Design, Automation and Test in Europe ConferenceDesign, Automation and Test in Europe Conference (DATE) 2022, ANTWERP, BELGIUM :},
institution = {Shahid Bahonar University of Kerman, Iran},
abstract = {Ternary Neural Networks (TNNs) compress network weights and activation functions into 2-bit representation resulting in remarkable network compression and energy efficiency. However, there remains a significant gap in accuracy between TNNs and full-precision counterparts. Recent advances in Neural Architectures Search (NAS) promise opportunities in automated optimization for various deep learning tasks. Unfortunately, this area is unexplored for optimizing TNNs. This paper proposes TAS, a framework that drastically reduces the accuracy gap between TNNs and their full-precision counterparts by integrating quantization into the network design. We experienced that directly applying NAS to the ternary domain provides accuracy degradation as the search settings are customized for full-precision networks. To address this problem, we propose (i) a new cell template for ternary networks with maximum gradient propagation; and (ii) a novel learnable quantizer that adaptively relaxes the ternarization mechanism from the distribution of the weights and activation functions. Experimental results reveal that TAS delivers 2.64% higher accuracy and 2.8x memory saving over competing methods with the same bit-width resolution on the CIFAR-10 dataset. These results suggest that TAS is an effective method that paves the way for the efficient design of the next generation of quantized neural networks.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Ternary Neural Networks (TNNs) compress network weights and activation functions into 2-bit representation resulting in remarkable network compression and energy efficiency. However, there remains a significant gap in accuracy between TNNs and full-precision counterparts. Recent advances in Neural Architectures Search (NAS) promise opportunities in automated optimization for various deep learning tasks. Unfortunately, this area is unexplored for optimizing TNNs. This paper proposes TAS, a framework that drastically reduces the accuracy gap between TNNs and their full-precision counterparts by integrating quantization into the network design. We experienced that directly applying NAS to the ternary domain provides accuracy degradation as the search settings are customized for full-precision networks. To address this problem, we propose (i) a new cell template for ternary networks with maximum gradient propagation; and (ii) a novel learnable quantizer that adaptively relaxes the ternarization mechanism from the distribution of the weights and activation functions. Experimental results reveal that TAS delivers 2.64% higher accuracy and 2.8x memory saving over competing methods with the same bit-width resolution on the CIFAR-10 dataset. These results suggest that TAS is an effective method that paves the way for the efficient design of the next generation of quantized neural networks.
37.
Ma, Zhiyuan; Yu, Wenting; Zhang, Peng; Huang, Zhi; Lin, Anni; Xia, Yan
LPI Radar Waveform Recognition Based on Neural Architecture Search Journal Article
In: Computational Intelligence and Neuroscience, vol. 2022, 0000.
@article{Ma2022,
title = {LPI Radar Waveform Recognition Based on Neural Architecture Search},
author = {Zhiyuan Ma and Wenting Yu and Peng Zhang and Zhi Huang and Anni Lin and Yan Xia},
url = {https://doi.org/10.1155/2022/4628481},
journal = {Computational Intelligence and Neuroscience},
volume = {2022},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
36.
Salam, Hanan; Manoranjan, Viswonathan; Jian, Jiang; Celiktutan, Oya
Learning Personalised Models for Automatic Self-Reported Personality Recognition Technical Report
0000.
@techreport{Salam2022,
title = {Learning Personalised Models for Automatic Self-Reported Personality Recognition},
author = {Hanan Salam and Viswonathan Manoranjan and Jiang Jian and Oya Celiktutan},
url = {https://nyuscholars.nyu.edu/ws/files/139000086/Proceedings_ICCV_2021_Understanding_Social_Behavior_in_Dyadic_and_Small_Group_Interactions_Challenge_8.pdf},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
35.
Dhanaraja, Mayur; Dob, Huyen; Nairb, Dinesh; Xub, Cong
Leveraging Tensor Methods in Neural Architecture Search for the automatic development of lightweight Convolutional Neural Networks Journal Article
In: 0000.
@article{dhanarajaleveraging,
title = {Leveraging Tensor Methods in Neural Architecture Search for the automatic development of lightweight Convolutional Neural Networks},
author = {Mayur Dhanaraja and Huyen Dob and Dinesh Nairb and Cong Xub},
url = {https://assets.amazon.science/34/29/484392a6450b8af9b7646fe6db60/leveraging-tensor-methods-in-neural-architecture-search-for-the-automatic-development-of-lightweight-convolutional-neural-networks.pdf},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
34.
Yang, Chengrun; Bender, Gabriel; Liu, Hanxiao; Kindermans, Pieter-Jan; Udell, Madeleine; Lu, Yifeng; Le, Quoc; Huang, Da
TabNAS: Rejection Sampling for Neural Architecture Search on Tabular Datasets Technical Report
0000.
@techreport{https://doi.org/10.48550/arxiv.2204.07615,
title = {TabNAS: Rejection Sampling for Neural Architecture Search on Tabular Datasets},
author = {Chengrun Yang and Gabriel Bender and Hanxiao Liu and Pieter-Jan Kindermans and Madeleine Udell and Yifeng Lu and Quoc Le and Da Huang},
url = {https://arxiv.org/abs/2204.07615},
doi = {10.48550/ARXIV.2204.07615},
publisher = {arXiv},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
33.
Hao, Yao; Zhang, Xizhe; Wang, Jie; Zhao, Tianyu; Sun, Baozhou
Improvement of IMRT QA prediction using imaging-based neural architecture search Journal Article
In: Medical Physics, vol. n/a, no. n/a, 0000.
@article{https://doi.org/10.1002/mp.15694,
title = {Improvement of IMRT QA prediction using imaging-based neural architecture search},
author = {Yao Hao and Xizhe Zhang and Jie Wang and Tianyu Zhao and Baozhou Sun},
url = {https://aapm.onlinelibrary.wiley.com/doi/abs/10.1002/mp.15694},
doi = {https://doi.org/10.1002/mp.15694},
journal = {Medical Physics},
volume = {n/a},
number = {n/a},
abstract = {Abstract Purpose Machine learning (ML) has been used to predict the gamma passing rate (GPR) of intensity-modulated radiation therapy (IMRT) QA results. In this work, we applied a novel neural architecture search to automatically tune and search for the best deep neural networks instead of using hand-designed deep learning architectures. Method and materials One hundred and eighty-two IMRT plans were created and delivered with portal dosimetry. A total of 1497 fields for multiple treatment sites were delivered and measured by portal imagers. Gamma criteria of 2%/2 mm with a 5% threshold were used. Fluence maps calculated for each plan were used as inputs to a convolution neural network (CNN). Auto-Keras was implemented to search for the best CNN architecture for fluence image regression. The network morphism was adopted in the searching process, in which the base models were ResNet and DenseNet. The performance of this CNN approach was compared with tree-based ML models previously developed for this application, using the same dataset. Results The deep-learning-based approach had 98.3% of predictions within 3% of the measured 2%/2-mm GPRs with a maximum error of 3.1% and a mean absolute error of less than 1%. Our results show that this novel architecture search approach achieves comparable performance to the machine-learning-based approaches with handcrafted features. Conclusions We implemented a novel CNN model using imaging-based neural architecture for IMRT QA prediction. The imaging-based deep-learning method does not require a manual extraction of relevant features and is able to automatically select the best network architecture.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Abstract Purpose Machine learning (ML) has been used to predict the gamma passing rate (GPR) of intensity-modulated radiation therapy (IMRT) QA results. In this work, we applied a novel neural architecture search to automatically tune and search for the best deep neural networks instead of using hand-designed deep learning architectures. Method and materials One hundred and eighty-two IMRT plans were created and delivered with portal dosimetry. A total of 1497 fields for multiple treatment sites were delivered and measured by portal imagers. Gamma criteria of 2%/2 mm with a 5% threshold were used. Fluence maps calculated for each plan were used as inputs to a convolution neural network (CNN). Auto-Keras was implemented to search for the best CNN architecture for fluence image regression. The network morphism was adopted in the searching process, in which the base models were ResNet and DenseNet. The performance of this CNN approach was compared with tree-based ML models previously developed for this application, using the same dataset. Results The deep-learning-based approach had 98.3% of predictions within 3% of the measured 2%/2-mm GPRs with a maximum error of 3.1% and a mean absolute error of less than 1%. Our results show that this novel architecture search approach achieves comparable performance to the machine-learning-based approaches with handcrafted features. Conclusions We implemented a novel CNN model using imaging-based neural architecture for IMRT QA prediction. The imaging-based deep-learning method does not require a manual extraction of relevant features and is able to automatically select the best network architecture.
32.
Zhai, Qihang; Li, Yan; Zhang, Zilin; Li, Yunjie; Wang, Shafei
Adaptive feature extraction and fine-grained modulation recognition of multi-function radar under small sample conditions Journal Article
In: IET Radar, Sonar & Navigation, vol. n/a, no. n/a, 0000.
@article{https://doi.org/10.1049/rsn2.12273,
title = {Adaptive feature extraction and fine-grained modulation recognition of multi-function radar under small sample conditions},
author = {Qihang Zhai and Yan Li and Zilin Zhang and Yunjie Li and Shafei Wang},
url = {https://ietresearch.onlinelibrary.wiley.com/doi/abs/10.1049/rsn2.12273},
doi = {https://doi.org/10.1049/rsn2.12273},
journal = {IET Radar, Sonar & Navigation},
volume = {n/a},
number = {n/a},
abstract = {Abstract Multi-function radars (MFRs) are sophisticated sensors with fine-grained modes, which modify their modulation types and parameters range generating various signals to fulfil different tasks, such as surveillance and tracking. In electromagnetic reconnaissance, recognition of MFR fine-grained modes can provide a basis for analysing strategies and reaction. With the limit of real applications, it is hard to obtain a large number of labelled samples for existing methods to learn the difference between categories. Therefore, it is essential to develop new methods to extract general knowledge of MFRs and identify modes with only a few samples. This paper proposes a few-shot learning (FSL) framework based on efficient neural architecture search (ENAS) with high robustness and portability, which designs a suitable network structure automated and quickly adapts to new environments. The experimental results show that the proposed method can still achieve excellent fine-grained modulation recognition performance (92.6%) under the condition of -6 dB signal-to-noise ratio (SNR), even when each class only provides one fixed-duration signal sample. The robustness is also verified under different conditions.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Abstract Multi-function radars (MFRs) are sophisticated sensors with fine-grained modes, which modify their modulation types and parameters range generating various signals to fulfil different tasks, such as surveillance and tracking. In electromagnetic reconnaissance, recognition of MFR fine-grained modes can provide a basis for analysing strategies and reaction. With the limit of real applications, it is hard to obtain a large number of labelled samples for existing methods to learn the difference between categories. Therefore, it is essential to develop new methods to extract general knowledge of MFRs and identify modes with only a few samples. This paper proposes a few-shot learning (FSL) framework based on efficient neural architecture search (ENAS) with high robustness and portability, which designs a suitable network structure automated and quickly adapts to new environments. The experimental results show that the proposed method can still achieve excellent fine-grained modulation recognition performance (92.6%) under the condition of -6 dB signal-to-noise ratio (SNR), even when each class only provides one fixed-duration signal sample. The robustness is also verified under different conditions.
31.
Wang, Ye-Qun; Li, Jian-Yu; Chen, Chun-Hua; Zhang, Jun; Zhan, Zhi-Hui
In: CAAI Transactions on Intelligence Technology, vol. n/a, no. n/a, 0000.
@article{https://doi.org/10.1049/cit2.12106,
title = {Scale adaptive fitness evaluation-based particle swarm optimisation for hyperparameter and architecture optimisation in neural networks and deep learning},
author = {Ye-Qun Wang and Jian-Yu Li and Chun-Hua Chen and Jun Zhang and Zhi-Hui Zhan},
url = {https://ietresearch.onlinelibrary.wiley.com/doi/abs/10.1049/cit2.12106},
doi = {https://doi.org/10.1049/cit2.12106},
journal = {CAAI Transactions on Intelligence Technology},
volume = {n/a},
number = {n/a},
abstract = {Abstract Research into automatically searching for an optimal neural network (NN) by optimisation algorithms is a significant research topic in deep learning and artificial intelligence. However, this is still challenging due to two issues: Both the hyperparameter and architecture should be optimised and the optimisation process is computationally expensive. To tackle these two issues, this paper focusses on solving the hyperparameter and architecture optimization problem for the NN and proposes a novel light-weight scale-adaptive fitness evaluation-based particle swarm optimisation (SAFE-PSO) approach. Firstly, the SAFE-PSO algorithm considers the hyperparameters and architectures together in the optimisation problem and therefore can find their optimal combination for the globally best NN. Secondly, the computational cost can be reduced by using multi-scale accuracy evaluation methods to evaluate candidates. Thirdly, a stagnation-based switch strategy is proposed to adaptively switch different evaluation methods to better balance the search performance and computational cost. The SAFE-PSO algorithm is tested on two widely used datasets: The 10-category (i.e., CIFAR10) and the 100−category (i.e., CIFAR100). The experimental results show that SAFE-PSO is very effective and efficient, which can not only find a promising NN automatically but also find a better NN than compared algorithms at the same computational cost.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Abstract Research into automatically searching for an optimal neural network (NN) by optimisation algorithms is a significant research topic in deep learning and artificial intelligence. However, this is still challenging due to two issues: Both the hyperparameter and architecture should be optimised and the optimisation process is computationally expensive. To tackle these two issues, this paper focusses on solving the hyperparameter and architecture optimization problem for the NN and proposes a novel light-weight scale-adaptive fitness evaluation-based particle swarm optimisation (SAFE-PSO) approach. Firstly, the SAFE-PSO algorithm considers the hyperparameters and architectures together in the optimisation problem and therefore can find their optimal combination for the globally best NN. Secondly, the computational cost can be reduced by using multi-scale accuracy evaluation methods to evaluate candidates. Thirdly, a stagnation-based switch strategy is proposed to adaptively switch different evaluation methods to better balance the search performance and computational cost. The SAFE-PSO algorithm is tested on two widely used datasets: The 10-category (i.e., CIFAR10) and the 100−category (i.e., CIFAR100). The experimental results show that SAFE-PSO is very effective and efficient, which can not only find a promising NN automatically but also find a better NN than compared algorithms at the same computational cost.
30.
Hu, Weifei; jinyi Shao,; Jiao, Qing; Wang, Chuxuan; Cheng, Jin; Liu, Zhenyu; Tan, Jianrong
A new differentiable architecture search method for optimizing convolutional neural networks in the digital twin of intelligent robotic grasping Journal Article
In: Journal of Intelligent Manufacturing, 0000.
@article{nokey,
title = {A new differentiable architecture search method for optimizing convolutional neural networks in the digital twin of intelligent robotic grasping},
author = {Weifei Hu and jinyi Shao and Qing Jiao and Chuxuan Wang and Jin Cheng and Zhenyu Liu and Jianrong Tan},
url = {https://link.springer.com/article/10.1007/s10845-022-01971-8},
journal = { Journal of Intelligent Manufacturing},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
29.
Xue, Fanghui
Relaxation and Optimization for Automated Learning of Neural Network Architectures DISSERTATION PhD Thesis
0000.
@phdthesis{XuePHD2022,
title = {Relaxation and Optimization for Automated Learning of Neural Network Architectures DISSERTATION},
author = {Fanghui Xue},
url = {https://escholarship.org/content/qt3wt239sm/qt3wt239sm.pdf},
keywords = {},
pubstate = {published},
tppubtype = {phdthesis}
}
28.
Mokhtari, Nassim; Nédélec, Alexis; Gilles, Marlene; Loor, Pierre De
Improving Neural Architecture Search by Mixing a FireFly algorithm with a Training Free Evaluation Technical Report
0000.
@techreport{mokhtariimproving,
title = {Improving Neural Architecture Search by Mixing a FireFly algorithm with a Training Free Evaluation},
author = {Nassim Mokhtari and Alexis Nédélec and Marlene Gilles and Pierre De Loor},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
27.
Duggal, Rahul
0000.
@phdthesis{DuggalPhD22,
title = {Robust Efficient Edge AI: New Principles and Frameworks for Empowering Artifical Intelligence on Edge Device ON EDGE DEVICES},
author = {Rahul Duggal},
url = {https://smartech.gatech.edu/bitstream/handle/1853/67315/DUGGAL-DISSERTATION-2022.pdf?sequence=1},
keywords = {},
pubstate = {published},
tppubtype = {phdthesis}
}
26.
Wen, Hao; Kang, Jingsu
Searching for Effective Neural Network Architectures for Heart Murmur Detection from Phonocardiogram Technical Report
0000.
@techreport{WenTS22,
title = {Searching for Effective Neural Network Architectures for Heart Murmur Detection from Phonocardiogram},
author = {Hao Wen and Jingsu Kang},
url = {https://cinc.org/2022/Program/accepted/130_Preprint.pdf},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
25.
Loni, Mohammad
Efficient Design of Scalable Deep Neural Networks for Resource-Constrained Edge Devices PhD Thesis
0000.
@phdthesis{LoniPhD,
title = {Efficient Design of Scalable Deep Neural Networks for Resource-Constrained Edge Devices},
author = {Mohammad Loni},
url = {https://www.diva-portal.org/smash/record.jsf?pid=diva2%3A1695852&dswid=-3791},
keywords = {},
pubstate = {published},
tppubtype = {phdthesis}
}
24.
Wenfeng Feng and, Xin Zhang; Song, Qiushuang; Sun, Guoying
Incoherence of Deep Isotropic Neural Networks increase their performance on Image Classification Technical Manual
0000.
@manual{nokey,
title = {Incoherence of Deep Isotropic Neural Networks increase their performance on Image Classification},
author = { Wenfeng Feng and, Xin Zhang and Qiushuang Song and Guoying Sun
},
url = {https://www.preprints.org/manuscript/202210.0092/v1},
keywords = {},
pubstate = {published},
tppubtype = {manual}
}
23.
Malkova, Aleksandra; Amini, Massih-Reza; Denis, Benoit; Villien, Christophe
Radio Map Reconstruction with Deep Neural Networks in a Weakly Labeled Learning Context with use of Heterogeneous Side Information Technical Manual
0000.
@manual{Malkova2022,
title = {Radio Map Reconstruction with Deep Neural Networks in a Weakly Labeled Learning Context with use of Heterogeneous Side Information},
author = {Aleksandra Malkova and Massih-Reza Amini and Benoit Denis and Christophe Villien},
url = {https://hal.archives-ouvertes.fr/hal-03823629/document},
keywords = {},
pubstate = {published},
tppubtype = {manual}
}
22.
Aboalam, Kawther; Neuswirth, Christoph; Pernau, Florian; Schiebel, Stefan; Spaethe, Fabian; Strohrmann, Manfred
Image Processing and Neural Network Optimization Methods for Automatic Visual Inspection Technical Manual
0000.
@manual{Aboalam2022,
title = {Image Processing and Neural Network Optimization Methods for Automatic Visual Inspection},
author = {Kawther Aboalam and Christoph Neuswirth and Florian Pernau and Stefan Schiebel and Fabian Spaethe and Manfred Strohrmann},
url = {https://www.researchgate.net/profile/Christoph-Reich/publication/364343172_Artificial_Intelligence_--_Applications_in_Medicine_and_Manufacturing_--_The_Upper_Rhine_Artificial_Intelligence_Symposium_UR-AI_2022/links/634cfa3476e39959d6c8bfb2/Artificial-Intelligence--Applications-in-Medicine-and-Manufacturing--The-Upper-Rhine-Artificial-Intelligence-Symposium-UR-AI-2022.pdf#page=33},
keywords = {},
pubstate = {published},
tppubtype = {manual}
}
21.
Mishra, Vidyanand; Kane, Lalit
A survey of designing convolutional neural network using evolutionary algorithms Journal Article
In: Artificial Intelligence Review, 0000.
@article{Mishra-AIR2022,
title = {A survey of designing convolutional neural network using evolutionary algorithms},
author = {Vidyanand Mishra and Lalit Kane
},
url = {https://link.springer.com/article/10.1007/s10462-022-10303-4},
journal = { Artificial Intelligence Review},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
20.
Sun, Yanan; Yen, Gary G.; Zhang, Mengjie
Evolutionary Deep Neural Architecture Search: Fundamentals, Methods, and recent AdvanceYa Book
0000.
@book{SunEDA22,
title = {Evolutionary Deep Neural Architecture Search: Fundamentals, Methods, and recent AdvanceYa},
author = {Yanan Sun and Gary G. Yen and Mengjie Zhang},
url = {https://books.google.de/books?hl=de&lr=&id=2RWbEAAAQBAJ&oi=fnd&pg=PR5&dq=%22neural+architecture+search%22&ots=yjnrR-vqyW&sig=0KFGVSnhQWTc1sQmWWewvmeuGqw#v=onepage&q=%22neural%20architecture%20search%22&f=false},
keywords = {},
pubstate = {published},
tppubtype = {book}
}
19.
Park, Minje
Proxy Data Generation for Fast and Efficient Neural Architecture Search Journal Article
In: Journal of Electrical Engineering & Technology 2022, 0000.
@article{ParkJEET22,
title = {Proxy Data Generation for Fast and Efficient Neural Architecture Search},
author = {
Minje Park
},
url = {https://link.springer.com/article/10.1007/s42835-022-01321-x},
journal = {Journal of Electrical Engineering & Technology 2022},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
18.
Singh, Yeshwant; Biswas, Anupam
In: Expert Systems, vol. n/a, no. n/a, pp. e13241, 0000.
@article{https://doi.org/10.1111/exsy.13241,
title = {Lightweight convolutional neural network architecture design for music genre classification using evolutionary stochastic hyperparameter selection},
author = {Yeshwant Singh and Anupam Biswas},
url = {https://onlinelibrary.wiley.com/doi/abs/10.1111/exsy.13241},
doi = {https://doi.org/10.1111/exsy.13241},
journal = {Expert Systems},
volume = {n/a},
number = {n/a},
pages = {e13241},
abstract = {Abstract Convolutional neural networks (CNNs) have succeeded in various domains, including music information retrieval (MIR). Music genre classification (MGC) is one such task in the MIR that has gained attention over the years because of the massive increase in online music content. Accurate indexing and automatic classification of these large volumes of music content require high computational resources, which pose a significant challenge to building a lightweight system. CNNs are a popular deep learning-based choice for building systems for MGC. However, finding an optimal CNN architecture for MGC requires domain knowledge both in CNN architecture design and music. We present MGA-CNN, a genetic algorithm-based approach with a novel stochastic hyperparameter selection for finding an optimal lightweight CNN-based architecture for the MGC task. The proposed approach is unique in automating the CNN architecture design for the MGC task. MGA-CNN is evaluated on three widely used music datasets and compared with seven peer rivals, which include three automatic CNN architecture design approaches and four manually designed popular CNN architectures. The experimental results show that MGA-CNN surpasses the peer approaches in terms of classification accuracy, parameter numbers, and execution time. The optimal architectures generated by MGA-CNN also achieve classification accuracy comparable to the manually designed CNN architectures while spending fewer computing resources.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Abstract Convolutional neural networks (CNNs) have succeeded in various domains, including music information retrieval (MIR). Music genre classification (MGC) is one such task in the MIR that has gained attention over the years because of the massive increase in online music content. Accurate indexing and automatic classification of these large volumes of music content require high computational resources, which pose a significant challenge to building a lightweight system. CNNs are a popular deep learning-based choice for building systems for MGC. However, finding an optimal CNN architecture for MGC requires domain knowledge both in CNN architecture design and music. We present MGA-CNN, a genetic algorithm-based approach with a novel stochastic hyperparameter selection for finding an optimal lightweight CNN-based architecture for the MGC task. The proposed approach is unique in automating the CNN architecture design for the MGC task. MGA-CNN is evaluated on three widely used music datasets and compared with seven peer rivals, which include three automatic CNN architecture design approaches and four manually designed popular CNN architectures. The experimental results show that MGA-CNN surpasses the peer approaches in terms of classification accuracy, parameter numbers, and execution time. The optimal architectures generated by MGA-CNN also achieve classification accuracy comparable to the manually designed CNN architectures while spending fewer computing resources.
17.
Gupta, Pritha; Drees, Jan Peter; Hüllermeier, Eyke
Automated Side-Channel Attacks using Black-Box Neural Architecture Search Technical Report
0000.
@techreport{Gupta22,
title = {Automated Side-Channel Attacks using Black-Box Neural Architecture Search},
author = {Pritha Gupta and Jan Peter Drees and Eyke Hüllermeier},
url = {https://eprint.iacr.org/2023/093.pdf},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
16.
Pandelea, Vlad; Ragusa, Edoardo; Gastaldo, Paolo; Cambria, Erik
SELECTING LANGUAGE MODELS FEATURES VIA SOFTWARE-HARDWARE CO-DESIGN Miscellaneous
0000.
@misc{Pandelea23,
title = {SELECTING LANGUAGE MODELS FEATURES VIA SOFTWARE-HARDWARE CO-DESIGN},
author = {Vlad Pandelea and Edoardo Ragusa and Paolo Gastaldo and Erik Cambria },
url = {https://w.sentic.net/selecting-language-models-features-via-software-hardware-co-design.pdf},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
15.
Huynh, Lam
FROM 3D SENSING TO DENSE PREDICTION PhD Thesis
0000.
@phdthesis{HuynhPhD23,
title = {FROM 3D SENSING TO DENSE PREDICTION},
author = {Lam Huynh},
url = {http://jultika.oulu.fi/files/isbn9789526235165.pdf},
keywords = {},
pubstate = {published},
tppubtype = {phdthesis}
}
14.
Cho, Minsu
Deep Learning Model Design Algorithms for High-Performing Plaintext and Ciphertext Inference PhD Thesis
0000.
@phdthesis{ChoPHD23,
title = {Deep Learning Model Design Algorithms for High-Performing Plaintext and Ciphertext Inference},
author = {Minsu Cho},
url = {https://www.proquest.com/docview/2767241424?pq-origsite=gscholar&fromopenview=true},
keywords = {},
pubstate = {published},
tppubtype = {phdthesis}
}
13.
Zhou, Dongzhan
Designing Deep Model and Training Paradigm for Object Perception PhD Thesis
0000.
@phdthesis{ZhouPhD2023,
title = {Designing Deep Model and Training Paradigm for Object Perception},
author = {Zhou, Dongzhan
},
url = {https://ses.library.usyd.edu.au/handle/2123/31055},
keywords = {},
pubstate = {published},
tppubtype = {phdthesis}
}
12.
Shariatzadeh, Seyed Mahdi; Fathy, Mahmood; Berangi, Reza
Improving the accuracy and speed of fast template-matching algorithms by neural architecture search Journal Article
In: Expert Systems, vol. n/a, no. n/a, pp. e13358, 0000.
@article{https://doi.org/10.1111/exsy.13358,
title = {Improving the accuracy and speed of fast template-matching algorithms by neural architecture search},
author = {Seyed Mahdi Shariatzadeh and Mahmood Fathy and Reza Berangi},
url = {https://onlinelibrary.wiley.com/doi/abs/10.1111/exsy.13358},
doi = {https://doi.org/10.1111/exsy.13358},
journal = {Expert Systems},
volume = {n/a},
number = {n/a},
pages = {e13358},
abstract = {Abstract Neural architecture search can be used to find convolutional neural architectures that are precise and robust while enjoying enough speed for industrial image processing applications. In this paper, our goal is to achieve optimal convolutional neural networks (CNNs) for multiple-templates matching for applications such as licence plates detection (LPD). We perform an iterative local neural architecture search for the models with minimum validation error as well as low computational cost from our search space of about 32 billion models. We describe the findings of the experience and discuss the specifications of the final optimal architectures. About 20-times error reduction and 6-times computational complexity reduction is achieved over our engineered neural architecture after about 500 neural architecture evaluation (in about 10 h). The typical speed of our final model is comparable to classic template matching algorithms while performing more robust and multiple-template matching with different scales.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Abstract Neural architecture search can be used to find convolutional neural architectures that are precise and robust while enjoying enough speed for industrial image processing applications. In this paper, our goal is to achieve optimal convolutional neural networks (CNNs) for multiple-templates matching for applications such as licence plates detection (LPD). We perform an iterative local neural architecture search for the models with minimum validation error as well as low computational cost from our search space of about 32 billion models. We describe the findings of the experience and discuss the specifications of the final optimal architectures. About 20-times error reduction and 6-times computational complexity reduction is achieved over our engineered neural architecture after about 500 neural architecture evaluation (in about 10 h). The typical speed of our final model is comparable to classic template matching algorithms while performing more robust and multiple-template matching with different scales.
11.
Yang, Yongjia; Zhan, Jinyu; Jiang, Wei; Jiang, Yucheng; Yu, Antai
Neural architecture search for resource constrained hardware devices: A survey Journal Article
In: IET Cyber-Physical Systems: Theory & Applications, vol. n/a, no. n/a, 0000.
@article{https://doi.org/10.1049/cps2.12058,
title = {Neural architecture search for resource constrained hardware devices: A survey},
author = {Yongjia Yang and Jinyu Zhan and Wei Jiang and Yucheng Jiang and Antai Yu},
url = {https://ietresearch.onlinelibrary.wiley.com/doi/abs/10.1049/cps2.12058},
doi = {https://doi.org/10.1049/cps2.12058},
journal = {IET Cyber-Physical Systems: Theory & Applications},
volume = {n/a},
number = {n/a},
abstract = {Abstract With the emergence of powerful and low-energy Internet of Things devices, deep learning computing is increasingly applied to resource-constrained edge devices. However, the mismatch between hardware devices with low computing capacity and the increasing complexity of Deep Neural Network models, as well as the growing real-time requirements, bring challenges to the design and deployment of deep learning models. For example, autonomous driving technologies rely on real-time object detection of the environment, which cannot tolerate the extra latency of sending data to the cloud, processing and then sending the results back to edge devices. Many studies aim to find innovative ways to reduce the size of deep learning models, the number of Floating-point Operations per Second, and the time overhead of inference. Neural Architecture Search (NAS) makes it possible to automatically generate efficient neural network models. The authors summarise the existing NAS methods on resource-constrained devices and categorise them according to single-objective or multi-objective optimisation. We review the search space, the search algorithm and the constraints of NAS on hardware devices. We also explore the challenges and open problems of hardware NAS.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Abstract With the emergence of powerful and low-energy Internet of Things devices, deep learning computing is increasingly applied to resource-constrained edge devices. However, the mismatch between hardware devices with low computing capacity and the increasing complexity of Deep Neural Network models, as well as the growing real-time requirements, bring challenges to the design and deployment of deep learning models. For example, autonomous driving technologies rely on real-time object detection of the environment, which cannot tolerate the extra latency of sending data to the cloud, processing and then sending the results back to edge devices. Many studies aim to find innovative ways to reduce the size of deep learning models, the number of Floating-point Operations per Second, and the time overhead of inference. Neural Architecture Search (NAS) makes it possible to automatically generate efficient neural network models. The authors summarise the existing NAS methods on resource-constrained devices and categorise them according to single-objective or multi-objective optimisation. We review the search space, the search algorithm and the constraints of NAS on hardware devices. We also explore the challenges and open problems of hardware NAS.
10.
Yan, Longhao; Wu, Qingyu; Li, Xi; Xie, Chenchen; Zhou, Xilin; Li, Yuqi; Shi, Daijing; Yu, Lianfeng; Zhang, Teng; Tao, Yaoyu; Yan, Bonan; Zhong, Min; Song, Zhitang; Yang, Yuchao; Huang, Ru
In: Advanced Functional Materials, vol. n/a, no. n/a, pp. 2300458, 0000.
@article{https://doi.org/10.1002/adfm.202300458,
title = {Neural Architecture Search with In-Memory Multiply–Accumulate and In-Memory Rank Based on Coating Layer Optimized C-Doped Ge2Sb2Te5 Phase Change Memory},
author = {Longhao Yan and Qingyu Wu and Xi Li and Chenchen Xie and Xilin Zhou and Yuqi Li and Daijing Shi and Lianfeng Yu and Teng Zhang and Yaoyu Tao and Bonan Yan and Min Zhong and Zhitang Song and Yuchao Yang and Ru Huang},
url = {https://onlinelibrary.wiley.com/doi/abs/10.1002/adfm.202300458},
doi = {https://doi.org/10.1002/adfm.202300458},
journal = {Advanced Functional Materials},
volume = {n/a},
number = {n/a},
pages = {2300458},
abstract = {Abstract Neural architecture search (NAS), as a subfield of automated machine learning, can design neural network models with better performance than manual design. However, the energy and time consumptions of conventional software-based NAS are huge, hindering its development and applications. Herein, 4 Mb phase change memory (PCM) chips are first fabricated that enable two key in-memory computing operations—in-memory multiply-accumulate (MAC) and in-memory rank for efficient NAS. The impacts of the coating layer material are systematically analyzed for the blade-type heating electrode on the device uniformity and in turn NAS performance. The random weights in the searched network architecture can be fine-tuned in the last stage. With 512 × 512 arrays based on 40 nm CMOS process, the PCM-based NAS has achieved 25–53× smaller model size and better performance than manually designed networks and improved the energy and time efficiency by 4779× and 123×, respectively, compared with NAS running on graphic processing unit (GPU). This work can expand the hardware accelerated in-memory operators, and significantly extend the applications of in-memory computing enabled by nonvolatile memory in advanced machine learning tasks.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Abstract Neural architecture search (NAS), as a subfield of automated machine learning, can design neural network models with better performance than manual design. However, the energy and time consumptions of conventional software-based NAS are huge, hindering its development and applications. Herein, 4 Mb phase change memory (PCM) chips are first fabricated that enable two key in-memory computing operations—in-memory multiply-accumulate (MAC) and in-memory rank for efficient NAS. The impacts of the coating layer material are systematically analyzed for the blade-type heating electrode on the device uniformity and in turn NAS performance. The random weights in the searched network architecture can be fine-tuned in the last stage. With 512 × 512 arrays based on 40 nm CMOS process, the PCM-based NAS has achieved 25–53× smaller model size and better performance than manually designed networks and improved the energy and time efficiency by 4779× and 123×, respectively, compared with NAS running on graphic processing unit (GPU). This work can expand the hardware accelerated in-memory operators, and significantly extend the applications of in-memory computing enabled by nonvolatile memory in advanced machine learning tasks.
9.
Addad, Youva; ad Frédéric Jurie, Alexis Lechervy
Multi-Exit Resource-Efficient Neural Architecture for Image Classification with Optimized Fusion Block Technical Report
0000.
@techreport{Addad-hal23a,
title = {Multi-Exit Resource-Efficient Neural Architecture for Image Classification with Optimized Fusion Block},
author = {Youva Addad and Alexis Lechervy ad Frédéric Jurie},
url = {https://hal.science/hal-04181149/document},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
8.
Tomczak, Nathaniel; Kuppannagari, Sanmukh
Automated Indexing Of TEM Diffraction Patterns Using Machine Learning Technical Report
0000.
@techreport{Tomczak-ieee-hpec23a,
title = {Automated Indexing Of TEM Diffraction Patterns Using Machine Learning},
author = {Nathaniel Tomczak and Sanmukh Kuppannagari},
url = {https://ieee-hpec.org/wp-content/uploads/2023/09/143.pdf},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
7.
Taccioli, Tommaso; Ragusa, Edoardo; Pomili, Tania; Gastaldo, Paolo; Pompa, Pier Paolo
Semi-quantitative determination of thiocyanate in saliva through colorimetric assays: design of CNN architecture via input-aware NAS Journal Article
In: IEEE SENSORS JOURNAL, , 0000.
@article{TaccioliSC23a,
title = {Semi-quantitative determination of thiocyanate in saliva through colorimetric assays: design of CNN architecture via input-aware NAS},
author = {Tommaso Taccioli and Edoardo Ragusa and Tania Pomili and Paolo Gastaldo and Pier Paolo Pompa},
url = {https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=10295395},
doi = {10.1109/JSEN.2023.3325545},
journal = {IEEE SENSORS JOURNAL, },
keywords = {},
pubstate = {published},
tppubtype = {article}
}
6.
Ahmed, Mohammed; Du, Hongbo; AlZoubi, Alaa
In: Ultrasonic Imaging, vol. 0, no. 0, pp. 01617346231208709, 0000, (PMID: 37981781).
@article{doi:10.1177/01617346231208709,
title = {ENAS-B: Combining ENAS With Bayesian Optimization for Automatic Design of Optimal CNN Architectures for Breast Lesion Classification From Ultrasound Images},
author = {Mohammed Ahmed and Hongbo Du and Alaa AlZoubi},
url = {https://doi.org/10.1177/01617346231208709},
doi = {10.1177/01617346231208709},
journal = {Ultrasonic Imaging},
volume = {0},
number = {0},
pages = {01617346231208709},
abstract = {Efficient Neural Architecture Search (ENAS) is a recent development in searching for optimal cell structures for Convolutional Neural Network (CNN) design. It has been successfully used in various applications including ultrasound image classification for breast lesions. However, the existing ENAS approach only optimizes cell structures rather than the whole CNN architecture nor its trainable hyperparameters. This paper presents a novel framework for automatic design of CNN architectures by combining strengths of ENAS and Bayesian Optimization in two-folds. Firstly, we use ENAS to search for optimal normal and reduction cells. Secondly, with the optimal cells and a suitable hyperparameter search space, we adopt Bayesian Optimization to find the optimal depth of the network and optimal configuration of the trainable hyperparameters. To test the validity of the proposed framework, a dataset of 1522 breast lesion ultrasound images is used for the searching and modeling. We then evaluate the robustness of the proposed approach by testing the optimized CNN model on three external datasets consisting of 727 benign and 506 malignant lesion images. We further compare the CNN model with the default ENAS-based CNN model, and then with CNN models based on the state-of-the-art architectures. The results (error rate of no more than 20.6% on internal tests and 17.3% on average of external tests) show that the proposed framework generates robust and light CNN models.},
note = {PMID: 37981781},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Efficient Neural Architecture Search (ENAS) is a recent development in searching for optimal cell structures for Convolutional Neural Network (CNN) design. It has been successfully used in various applications including ultrasound image classification for breast lesions. However, the existing ENAS approach only optimizes cell structures rather than the whole CNN architecture nor its trainable hyperparameters. This paper presents a novel framework for automatic design of CNN architectures by combining strengths of ENAS and Bayesian Optimization in two-folds. Firstly, we use ENAS to search for optimal normal and reduction cells. Secondly, with the optimal cells and a suitable hyperparameter search space, we adopt Bayesian Optimization to find the optimal depth of the network and optimal configuration of the trainable hyperparameters. To test the validity of the proposed framework, a dataset of 1522 breast lesion ultrasound images is used for the searching and modeling. We then evaluate the robustness of the proposed approach by testing the optimized CNN model on three external datasets consisting of 727 benign and 506 malignant lesion images. We further compare the CNN model with the default ENAS-based CNN model, and then with CNN models based on the state-of-the-art architectures. The results (error rate of no more than 20.6% on internal tests and 17.3% on average of external tests) show that the proposed framework generates robust and light CNN models.
5.
Herron, Emily J.
0000.
@phdthesis{Herron-phd23a,
title = {Generalized Differentiable Neural Architecture Search with Performance and Stability ImprovementsPerformance and Stability Improvements},
author = {Emily J. Herron},
url = {https://trace.tennessee.edu/cgi/viewcontent.cgi?article=10188&context=utk_graddiss},
keywords = {},
pubstate = {published},
tppubtype = {phdthesis}
}
4.
Chen, Xiangning
Advancing Automated Machine Learning: Neural Architectures and Optimization Algorithms PhD Thesis
0000.
@phdthesis{Chen-phd23a,
title = {Advancing Automated Machine Learning: Neural Architectures and Optimization Algorithms},
author = {Chen, Xiangning},
url = {https://escholarship.org/content/qt2f40c1w4/qt2f40c1w4.pdf},
keywords = {},
pubstate = {published},
tppubtype = {phdthesis}
}
3.
Xiao, Songyi; Wang, Wenjun
Ranking-based architecture generation for surrogate-assisted neural architecture search Journal Article
In: Concurrency and Computation: Practice and Experience, vol. n/a, no. n/a, pp. e8051, 0000.
@article{https://doi.org/10.1002/cpe.8051,
title = {Ranking-based architecture generation for surrogate-assisted neural architecture search},
author = {Songyi Xiao and Wenjun Wang},
url = {https://onlinelibrary.wiley.com/doi/abs/10.1002/cpe.8051},
doi = {https://doi.org/10.1002/cpe.8051},
journal = {Concurrency and Computation: Practice and Experience},
volume = {n/a},
number = {n/a},
pages = {e8051},
abstract = {Abstract Architectures generation optimization has been received a lot of attention in neural architecture search (NAS) since its efficiency in generating architecture. By learning the architecture representation through unsupervised learning and constructing a latent space, the prediction process of predictors is simplified, leading to improved efficiency in architecture search. However, searching for architectures with top performance in complex and large NAS search spaces remains challenging. In this paper, an approach that combined a ranker and generative model is proposed to address this challenge through regularizing the latent space and identifying architectures with top rankings. We introduce the ranking error to gradually regulate the training of the generative model, making it easier to identify architecture representations in the latent space. Additionally, a surrogate-assisted evolutionary search method that utilized neural network Bayesian optimization is proposed to efficiently explore promising architectures in the latent space. We demonstrate the benefits of our approach in optimizing architectures with top rankings, and our method outperforms state-of-the-art techniques on various NAS benchmarks. The code is available at https://github.com/outofstyle/RAGS-NAS.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Abstract Architectures generation optimization has been received a lot of attention in neural architecture search (NAS) since its efficiency in generating architecture. By learning the architecture representation through unsupervised learning and constructing a latent space, the prediction process of predictors is simplified, leading to improved efficiency in architecture search. However, searching for architectures with top performance in complex and large NAS search spaces remains challenging. In this paper, an approach that combined a ranker and generative model is proposed to address this challenge through regularizing the latent space and identifying architectures with top rankings. We introduce the ranking error to gradually regulate the training of the generative model, making it easier to identify architecture representations in the latent space. Additionally, a surrogate-assisted evolutionary search method that utilized neural network Bayesian optimization is proposed to efficiently explore promising architectures in the latent space. We demonstrate the benefits of our approach in optimizing architectures with top rankings, and our method outperforms state-of-the-art techniques on various NAS benchmarks. The code is available at https://github.com/outofstyle/RAGS-NAS.
2.
Avval, Sasan Salmani Pour; Yaghoubi, Vahid; Eskue, Nathan D.; Groves, Roger M.
Systematic Review on Neural Architecture Search Technical Report
0000.
@techreport{Avval-24a,
title = {Systematic Review on Neural Architecture Search},
author = {Sasan Salmani Pour Avval and Vahid Yaghoubi and Nathan D. Eskue and Roger M. Groves},
url = {https://www.researchsquare.com/article/rs-4085293/v1},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
1.
Yao, Yang; Wang, Xin; Qin, Yijian; Zhang, Ziwei; Zhu, Wenwu; Mei, Hong
Customized Cross-device Neural Architecture Search with Images Miscellaneous
0000.
@misc{Yao2024,
title = {Customized Cross-device Neural Architecture Search with Images},
author = {Yang Yao and Xin Wang and Yijian Qin and Ziwei Zhang and Wenwu Zhu and Hong Mei},
url = {http://mn.cs.tsinghua.edu.cn/xinwang/PDF/papers/2024_Customized%20Cross-device%20Neural%20Architecture%20Search%20with%20Images.pdf},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}