检索结果-内蒙古大学图书馆

The Journal of Machine Learning Research 2024年第1期25卷 9702-9734页

作者： Julien Keisler El-Ghazali Talbi Sandra Claudel Gilles Cabriel EDF Lab Paris-Saclay Palaiseau and University of Lille & INRIA Lille University of Lille & INRIA Lille EDF Lab Paris-Saclay Palaiseau

In this paper, we propose DRAGON (for DiRected Acyclic Graph optimization), an algorithmic framework to automatically generate efficient deep neural networks architectures and optimize their associated hyperparameters. The framework is based on evolving Directed Acyclic Graphs (DAGs), defining a more exible search space than the existing ones in the literature. It allows mixtures of different classical operations: convolutions, recurrences and dense layers, but also more newfangled operations such as self-attention. Based on this search space we propose neighbourhood and evolution search operators to optimize both the architecture and hyper-parameters of our networks. These search operators can be used with any metaheuristic capable of handling mixed search spaces. We tested our algorithmic framework with an asynchronous evolutionary algorithm on a time series forecasting benchmark. The results demonstrate that DRAGON outperforms state-of-the-art handcrafted models and AutoML techniques for time series forecasting on numerous datasets. DRAGON has been implemented as a python open-source package1.

关键词： neural architecture search hyperparameters optimization metaheuristics evolutionary algorithm time series forecasting

来源：评论

学校读者我要写书评

暂无评论

optimization and application of the distribution of relaxation times based on characteristic frequency resolution and hyperparameters

引用

JOURNAL OF POWER SOURCES 2022年 545卷

作者： Zhu, Dong Ma, Tiancai Yang, Yanbo Tongji Univ Sch Automot Studies Shanghai 201804 Peoples R China

Distribution of relaxation times (DRT) is a promising method for interpreting electrochemical impedance spectroscopy (EIS) data. Contemporary studies have made significant progress in theory, algorithm, imple-mentation, and application of DRT. However, the DRT deconvolution is often difficult to interpret because of the pseudo-peaks, peaks deviation or peaks overlapping. In this paper, a comprehensive regularization DRT method based on characteristic frequency resolution and hyperparameters is proposed. Based on quantitative analysis of synthetic simulation data, the characteristic frequency resolution optimization is applied to avoid the peaks overlapping and improve the accuracy of DRT. Meanwhile, the hyperparameters optimization which assigns weights to each individual data point is verified to improve the robustness of DRT. Then, the optimized DRT is applied to the EIS experimental data analysis of proton exchange membrane fuel cell (PEMFC), and the polar-ization processes of PEMFC are identified. Moreover, a fast EIS measurement based on characteristic frequency of DRT is proposed and verified by the impedance prediction which the deviations are within 10%, but the EIS measurement time is shortened from 6 min to 5 s. Moreover, this method can provide a new direction and idea for fault diagnosis based on EIS data, especially for on-board applications.

关键词： Distribution of relaxation times Electrochemical impedance spectroscopy Characteristic frequency resolution hyperparameters optimization Proton exchange membrane fuel cell

来源：评论

学校读者我要写书评

暂无评论

Stall prediction model based on deep learning network in axial flow compressor

引用

CHINESE JOURNAL OF AERONAUTICS 2025年第4期38卷 44-54页

作者： Deng, Yuyang Li, Jichao Liu, Jingyuan Peng, Feng Zhang, Hongwu Schoen, Marco P. Chinese Acad Sci Adv Gas Turbine Lab Inst Engn Thermophys Beijing 100190 Peoples R China Univ Chinese Acad Sci Sch Aeronaut & Astronaut Beijing 100049 Peoples R China Natl Key Lab Sci & Technol Adv Light duty Gas Turb Beijing 100049 Peoples R China Chinese Acad Sci Inst Engn Thermophys Key Lab Adv Energy & Power Beijing 100190 Peoples R China Jiangsu Univ Res Ctr Fluid Machinery Engn & Technol Zhenjiang 212000 Peoples R China Idaho State Univ Dept Mech Engn Pocatello ID 83209 USA

To predict stall and surge in advance that make the aero-engine compressor operate safely, a stall prediction model based on deep learning theory is established in the current study. The Long Short-Term Memory (LSTM) originating from the recurrent neural network is used, and a set of measured dynamic pressure datasets including the stall process is used to learn what determines the weight of neural network nodes. Subsequently, the structure and function hyperparameters in the model are deeply optimized, and a set of measured pressure data is used to verify the prediction effects of the model. On this basis of the above good predictive capability, stall in lowand high-speed compressor are predicted by using the established model. When a period of non-stall pressure data is used as input in the model, the model can quickly complete the prediction of subsequent time series data through the self-learning and prediction mechanism. Comparison with the real-time measured pressure data demonstrates that the starting point of the predicted stall is basically the same as that of the measured stall, and the stall can be predicted more than 1 s in advance so that the occurrence of stall can be avoided. The model of stall prediction in the current study can make up for the uncertainty of threshold selection of the existing stall warning methods based on measured data signal processing. It has a great application potential to predict the stall occurrence of aero-engine compressor in advance and avoid the accidents. (c) 2024 Published by Elsevier Ltd on behalf of Chinese Society of Aeronautics and Astronautics. This is an open access article under the CC BY-NC-ND license (http://***/licenses/by-nc-nd/4.0/).

关键词： Compressor Stall Deep learning LSTM hyperparameters optimization

来源：评论

学校读者我要写书评

暂无评论

MoAR-CNN: Multi-Objective Adversarially Robust Convolutional Neural Network for SAR Image Classification

引用

IEEE TRANSACTIONS ON EMERGING TOPICS IN COMPUTATIONAL INTELLIGENCE 2025年第1期9卷 57-74页

作者： Wei, Hai-Nan Zeng, Guo-Qiang Lu, Kang-Di Geng, Guang-Gang Weng, Jian Jinan Univ Coll Cyber Secur Guangzhou 510632 Peoples R China Jinan Univ Natl Joint Engn Res Ctr Network Secur Detect & Pro Guangzhou 510632 Peoples R China Wenzhou Univ Natl Local Joint Engn Res Ctr Digitalized Elect De Wenzhou 325035 Peoples R China Donghua Univ Coll Informat Sci & Technol Shanghai 201620 Peoples R China Zhejiang Univ Inst Cyber Syst & Control Hangzhou Peoples R China

Deep neural networks (DNNs) have been widely applied to the synthetic aperture radar (SAR) images detection and classification recently while different kinds of adversarial attacks from malicious adversary and the hidden vulnerability of DNNs may lead to serious security threats. The state-of-the-art DNNs-based SAR image detection models are designed manually by only considering the test accuracy performance on clean datasets but neglecting the models' adversarial robustness under various types of adversarial attacks. In order to obtain the best trade-off between the clean accuracy and adversarial robustness in robust convolutional neural networks (CNNs)-based SAR image classification models, this work makes the first attempt to develop a multi-objective adversarially robust CNN, called MoAR-CNN. In the MoAR-CNN, we propose a multi-objective automatic design method of the cells-based neural architectures and some critical hyperparameters such as the optimizer type and learning rate of CNNs. A Squeeze-and-Excitation (SE) layer is introduced after each cell to improve the computational efficiency and robustness. The experiments on FUSAR-Ship and OpenSARShip datasets against seven types of adversarial attacks have been implemented to demonstrate the superiority of the proposed MoAR-CNN to six classical manually designed CNNs and four robust neural architectures search methods in terms of clean accuracy, adversarial accuracy, and model size. Furthermore, we also demonstrate the advantages of using SE layer in MoAR-CNN, the transferability of MoAR-CNN, search costs, adversarial training, and the developed NSGA-II in MoAR-CNN through experiments.

关键词： Synthetic aperture radar (SAR) adversarial robustness convolutional neural networks multi-objective evolutionary algorithms hyperparameters optimization neural architectures search Synthetic aperture radar (SAR) adversarial robustness convolutional neural networks multi-objective evolutionary algorithms hyperparameters optimization neural architectures search

来源：评论

学校读者我要写书评

暂无评论

Displacement response reconstruction method for bridge subjected to moving load based on IPSO-BiLSTM network

引用

SMART STRUCTURES AND SYSTEMS 2025年第3期35卷 141-151页

作者： He, Wen-Yu Gao, Ao Li, Yi-Fan Hu, Dong-Yang Hefei Univ Technol Hefei 230009 Anhui Peoples R China Anhui Prov Rd & Bridge Inspect Engn Res Ctr Hefei 230009 Anhui Peoples R China Kunming Survey Design & Res Inst Co Ltd CREEC Kunming 650200 Peoples R China

Bridge dynamic displacement reconstruction methods based on neural networks usually use single-input neural networks, and most of the hyperparameters are determined by experience, which seriously affect the reconstruction accuracy. In this paper, a reconstruction method for bridge displacement response induced by moving load is proposed by using a small number of sensors and a triple-input IPSO-BiLSTM network. Firstly, the input strain and acceleration data are normalized in advance for data fusion. Secondly, IPSO-BILSTM network model with three-time sequence responses as input is constructed, and IPSO algorithm is used to optimize the network hyperparameters. Finally, three-time sequence responses are input into the trained iterative particle swarm optimization (IPSO)-Bidirectional LSTM (IPSO-BiLSTM) neural network to reconstruct the bridge displacement response. The proposed IPSO-BiLSTM network realizes the data fusion of three-time sequence responses and automatically establishes the relationship between input response and output displacement. Numerical examples indicate that the reconstruction accuracy is sensitive to road roughness and measurement noise. Experimental studies reveal that the reconstruction accuracy is insensitive to vehicle velocity and weigh.

关键词： bridge displacement reconstruction hyperparameters optimization IPSO-BiLSTM Network moving load

来源：评论

学校读者我要写书评

暂无评论

An optimized deep hybrid learning for multi-channel EEG-based driver drowsiness detection

引用

BIOMEDICAL SIGNAL PROCESSING AND CONTROL 2025年 99卷

作者： Latreche, Imene Slatnia, Sihem Kazar, Okba Harous, Saad Univ Biskra Dept Comp Sci Biskra Algeria Univ Kalba Coll Arts Sci & Informat Technol Dept Comp Sci Sharjah U Arab Emirates Univ Sharjah Coll Comp & Informat Dept Comp Sci Sharjah U Arab Emirates

Driver drowsiness is a severe issue that has contributed to numerous fatal accidents and injuries. Thus, detecting driver drowsiness is an important task that has been the subject of intensive research in recent years. There have been numerous proposed physiological signals for detecting driver drowsiness. However, the Electroencephalographic (EEG) signal is the most commonly used due to its direct relationship with drowsiness and its simplicity of acquisition. Recently, different Machine Learning (ML) and Deep Learning (DL) models have been proposed to detect driver drowsiness. This study utilized a publicly accessible dataset containing twelve healthy participants. Reading numerous research papers, we determined no specific EEG-based drowsiness preprocessing parameter values. Consequently, as a first step, and for the first time, to our knowledge in this field, we applied an optimization algorithm to determine the optimal preprocessing parameter values using a CNN model and accuracy as the objective function. The obtained results demonstrated the importance of selecting the correct values, as the mean accuracy score increased from 91% before optimization to 95% after optimization for the proposed CNN. The training time has been reduced. Also, as a second step, we have used the Optuna Hyperparameter optimization framework to select the optimal CNN hyperparameters, which increased the mean accuracy from 95% to 97%. Finally, to take advantage of Deep Hybrid learning, we have fused the CNN "features extractor" with seven ML classifiers, with the CNN-SVM classifier achieving the highest average accuracy of 99.9%, and the training time has been reduced to a shallow value.

关键词： Electroencephalograph (EEG) Driver drowsiness Deep learning Machine learning Preprocessing optimization hyperparameters optimization Optuna Deep hybrid learning

来源：评论

学校读者我要写书评

暂无评论

Graph Structure Learning With Automatic Search of hyperparameters Based on Genetic Programming

引用

IEEE TRANSACTIONS ON EMERGING TOPICS IN COMPUTATIONAL INTELLIGENCE 2024年第6期8卷 4155-4164页

作者： Wang, Pengda Lu, Mingjie Yan, Weiqing Yang, Dong Liu, Zhaowei Univ Sci & Technol China Hefei 230026 Anhui Peoples R China Yantai Univ Sch Comp & Control Engn Yantai 264005 Peoples R China Georgia State Univ Dept Comp Sci Atlanta GA 30303 USA

Graph neural networks (GNNs) rely heavily on graph structures and artificial hyperparameters, which may increase computation and affect performance. Most GNNs use original graphs, but the original graph data has problems with noise and incomplete information, which easily leads to poor GNN performance. For this kind of problem, recent graph structure learning methods consider how to generate graph structures containing label information. The settings of some hyperparameters will also affect the expression of the GNN model. This paper proposes a genetic graph structure learning method (Genetic-GSL). Different from the existing graph structure learning methods, this paper not only optimizes the graph structure but also the hyperparameters. Specifically, different graph structures and different hyperparameters are used as parents;the offspring are cross-mutated through the parents;and then excellent offspring are selected through evaluation to achieve dynamic fitting of the graph structure and hyperparameters. Experiments show that, compared with other methods, Genetic-GSL basically improves the performance of node classification tasks by 1.2%. With the increase in evolution algebra, Genetic-GSL has good performance on node classification tasks and resistance to adversarial attacks.

关键词： Graph neural networks hyperparameters optimization graph structure learning genetic model

来源：评论

学校读者我要写书评

暂无评论

An enhanced deep learning model for high-speed classification of plant diseases with bioinspired algorithm

引用

JOURNAL OF SUPERCOMPUTING 2024年第3期80卷 3713-3737页

作者： Ahila, A. Prema, V. Ayyasamy, S. Sivasubramanian, M. Sethu Inst Technol Dept Elect & Commun Engn Virudunagar 626115 Tamil Nadu India JP Coll Arts & Sci Dept Comp Applicat Tenkasi 627852 Tamil Nadu India Vellore Inst Technol Sch Comp Sci & Engn Vellore Campus Vellore 632014 Tamilnadu India JP Coll Arts & Sci Dept Comp Sci Tenkasi 627852 Tamil Nadu India

Agriculture is one of the most crucial aspects of a nation's growth. However, the quality and quantity of crop yield are severely affected by various plant diseases. Plant diseases must be identified and prevented at an early stage to improve food quality and production rate. The emerging deep learning network of convolutional neural networks (CNNs) achieved excellent results in plant disease classification. However, the classification potential of the network depends largely on the configuration of hyperparameters. Finding the optimal set of hyperparameters is a tedious, time-consuming, and challenging task. To tackle such an issue, this paper proposes an optimized CNN integrated with a novel modified whale optimization algorithm (MWOA) to achieve plant disease classification. Here, the novel method of optimizing CNN hyperparameters facilitates quicker implementation. To boost the proposed model's efficiency, several data augmentation methods are used such as rotation, scaling, and flipping. The proposed MWOA-CNN model is implemented using a plant village database that includes fourteen plant species, 38 disease classes, and healthy leaves as well. The experiential findings showed that the proposed model outperforms the existing models by attaining the highest classification accuracy of 99.92%, resulting in an effective model for plant disease classification.

关键词： Deep learning hyperparameters optimization Image augmentation Plant disease classification Modified whale optimization algorithm

来源：评论

学校读者我要写书评

暂无评论

Heuristic based federated learning with adaptive hyperparameter tuning for households energy prediction

引用

SCIENTIFIC REPORTS 2025年第1期15卷 1-17页

作者： Toderean, Liana Daian, Mihai Cioara, Tudor Anghel, Ionut Michalakopoulos, Vasilis Sarantinopoulos, Efstathios Sarmas, Elissaios Tech Univ Cluj Napoca Comp Sci Dept Distributed Syst Res Lab G Baritiu 26-28 Cluj Napoca 400027 Romania Natl Tech Univ Athens Sch Elect & Comp Engn Decis Support Syst Lab Ir Politech 9 Athens 15773 Greece

Federated Learning is transforming electrical load forecasting by enabling Artificial Intelligence (AI) models to be trained directly on household edge devices. However, the prediction accuracy of federated learning models tends to diminish when dealing with non-IID data highlighting the need for adaptive hyperparameter optimization strategies to improve performance. In this paper, we propose a novel hierarchical federated learning solution for efficient model aggregation and hyperparameter tuning, specifically tailored to household energy prediction. The households with similar energy profiles are clustered at the edge, linked, and aggregated at the fog level, to enable effective and adaptive hyperparameter tuning. The federated model aggregation is optimized using hierarchical simulated annealing optimization to prioritize updates from the better-performing models. A genetic algorithm-based hyperparameter optimization method reduces the computational load on edge nodes by efficiently exploring different configurations and using only the most promising ones for edge nodes' cross-validation. The evaluation results demonstrate a significant improvement in average prediction accuracy and better capturing of energy patterns compared to the federated averaging approach. The impact on network traffic among nodes across different layers is kept below 30 KB. Additionally, hyperparameter tuning reduces the size of model updates and the number of communication rounds by 30%, which is particularly beneficial when network resources are limited.

关键词： Federated learning Energy prediction hyperparameters optimization Simulated Annealing-based federated aggregation Genetic algorithm

来源：评论

学校读者我要写书评

暂无评论

TL-LFF Net: transfer learning based lighter, faster, and frozen network for the detection of multi-scale mixed intracranial hemorrhages through genetic optimization algorithm

引用

INTERNATIONAL JOURNAL OF MACHINE LEARNING AND CYBERNETICS 2024年 1-27页

作者： Kothala, Lakshmi Prasanna Guntur, Sitaramanjaneya Reddy Vignans Fdn Sci Technol & Res Dept Elect & Commun Engn Guntur 522213 Andhra Prades India Vignans Fdn Sci Technol & Res Dept Biomed Engn Guntur 522213 Andhra Prades India

Computed tomography (CT) is the most commonly used imaging method in intracranial hemorrhage (ICH). Although deep learning (DL) models are well suited for detecting and segmenting multi-class hemorrhages, localizing multi-scale mixed hemorrhages with limited resources such as bounding boxes is difficult. To address this issue, the current study proposes a novel transfer learning-based TL-LFF Network. To detect multi-scale mixed hemorrhages, the proposed model employs a backbone module that extracts in-depth features from the input images, and a spatial pyramid pooling faster layer that performs the pooling operation at various levels. In the neck section, a path aggregated network (PANet) is used to store spatial information. Furthermore, to achieve a lightweight nature, the proposed backbone and neck modules were frozen during the backpropagation stage, resulting in a decrease in detection accuracy. To improve detection capability while remaining lightweight, a concept known as transfer learning is used. This strategy significantly improves the accuracy of the proposed model. In addition, the Genetic Algorithm (GA) concept is used to optimize the hyperparameters, where the mutation is used to develop new offspring based on previous generations. The brain hemorrhage extended dataset was used to train and validate the proposed model. In terms of detection metrics and lightweight criteria, the experimental results showed that the proposed model performed better when compared to other existing models. As a result, we can use the proposed model in the clinical implementation stage to reduce the radiologist's CT scan read time.

关键词： Deep learning Frozen YOLO Genetic algorithm hyperparameters optimization Mixed intracranial hemorrhage Transfer learning

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：