检索结果-内蒙古大学图书馆

Spatial attention-based hybrid VGG-SVM and VGG-RF frameworks for improved cotton leaf disease detection

Neural Computing and Applications 2025年第14期37卷 8309-8329页

作者： Pandiyaraju, V. Anusha, B. Senthil Kumar, A.M. Jaspin, K. Venkatraman, Shravan Kannan, A. School of Computer Science and Engineering Vellore Institute of Technology Tamil Nadu Chennai India Department of Computer Science and Engineering Marian Engineering College Trivandram Kerala Trivandrum India Department of Computer Science and Engineering St. Joseph’s Institute of Technology Tamil Nadu Chennai India Department of Information Science and Technology College of Engineering Guindy Anna University Tamil Nadu Chennai India

The agriculture industry's production and food quality have been impacted by plant leaf diseases in recent years. Hence, it is vital to have a system that can automatically identify and diagnose diseases at an initial stage for enhancing the quality of agricultural output and also for stopping the overall plant extinction. Most state-of-the-art methods do not provide appropriate accuracy of identification when the similarity among different diseases is higher and the input leaf images have complex background. In this work, a transfer learning-based automated crop disease recognition system is proposed to handle this problem. The cotton leaf disease detection framework being proposed here includes two key stages: (1) the feature extraction stage, where the convolutional neural network (CNN) model from the visual geometry group (VGG-16) is employed to obtain the deep features from cotton leaf images with complex background. (2) Cotton leaf disease detection stage: Here, the extracted optimal features are input to the machine learning classifiers, namely random forest (RF) and nonlinear support vector machine (SVM), for the initial classification of the diseases. For providing an optimal cotton crop disease detection performance, Kaggle’s four-class cotton disease dataset with complex background is chosen in this work. For the final decision making and analysis, two hybrid models named spatial attention-based hybrid VGG-RF and hybrid VGG-SVM have been proposed in this work for enhancing the decision accuracy. These hybrid models offered an average accuracy of approximately 98.29% and 99.31%, respectively, which is higher than the accuracy provided by the related work. This proved that the proposed hybrid model, namely spatial attention-based VGG-SVM and VGG-RF models, with spatial attention outperformed the state-of-the-art classifiers. © The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2025.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

HViTML: Hybrid vision transformer with machine learning-based classification model for glaucomatous eye

引用

Multimedia Tools and Applications 2025年 1-24页

作者： Singh, Piyush Bhushan Singh, Pawan Dev, Harsh Batra, Devanshu Chaurasia, Brijesh Kumar Department of Information Technology Pranveer Singh Institute of Technology Uttar Pradesh Kanpur India Department of Computer Science and Engineering Amity School of Engineering and Technology Amity University Uttar Pradesh Lucknow India Department of Computer Science and Engineering Babu Banarasi Das University Uttar Pradesh Lucknow India Department of Computer Science and Engineering Pranveer Singh Institute of Technology Uttar Pradesh Kanpur India

Glaucoma is currently one of the most significant causes of permanent blindness. Fundus imaging is the most popular glaucoma screening method because of the compromises it has to make in terms of portability, size, and cost. In recent years, convolution neural networks (CNNs) have revolutionized computer vision. Convolution is a "local" CNN technique that is only applicable to a small region surrounding an image. Vision Transformers (ViT) use self-attention, which is a "global" activity since it collects information from the entire image. As a result, the ViT can successfully gather distant semantic relevance from an image. This study examined several optimizers, including Adamax, SGD, RMSprop, Adadelta, Adafactor, Nadam, and Adagrad. With 1750 Healthy and Glaucoma images in the IEEE fundus image dataset and 4800 healthy and glaucoma images in the LAG fundus image dataset, we trained and tested the ViT model on these datasets. Additionally, the datasets underwent image scaling, auto-rotation, and auto-contrast adjustment via adaptive equalization during preprocessing. The results demonstrated that preparing the provided dataset with various optimizers improved accuracy and other performance metrics. Additionally, according to the results, the Nadam Optimizer improved accuracy in the adaptive equalized preprocessing of the IEEE dataset by up to 97.8% and in the adaptive equalized preprocessing of the LAG dataset by up to 92%, both of which were followed by auto rotation and image resizing processes. In addition to integrating our vision transformer model with the shift tokenization model, we also combined ViT with a hybrid model that consisted of six different models, including SVM, Gaussian NB, Bernoulli NB, Decision Tree, KNN, and Random Forest, based on which optimizer was the most successful for each dataset. Empirical results show that the SVM Model worked well and improved accuracy by up to 93% with precision of up to 94% in the adaptive equalization preprocess

关键词： Decision trees

来源：评论

学校读者我要写书评

暂无评论

Review on Meta-heuristic Algorithm-Based Priority-Aware Computation Offloading in Edge Computing System

引用

Journal of The Institution of Engineers (India): Series B 2025年 1-26页

作者： Pilli, Neelima Mohapatra, Debasis Reddy, Shiva Shankar Department of Computer Science and Engineering Biju Patnaik University of Technology Odisha Rourkela India Department of Computer Science and Engineering Parala Maharaja Engineering College Odisha Brahmpur India Department of Computer Science and Engineering Sagi Rama Krishnam Raju Engineering College Andhra Pradesh Bhimavaram India

Advancements in smart applications highlight the need for increased processing and storage capacity at Smart Devices (SDs). To tackle this, Edge computing (EC) is enabled to offload SD workloads to distant edge servers. So, a variety of offloading strategies are offered in previous research to enhance the efficiency of SDs by offloading their workloads to neighboring cloudlets or distant cloud computing resources. Consequently, optimization techniques are crucial for scheduling computation offloading tasks in (ED) networks. To the greatest extent of current understanding, an extensive analysis of meta-heuristic-based computation offloading (CO) techniques in the EC has not been systematically surveyed. This systematic survey creates an inventory of CO in EC and further summarizes the most advanced recently available EC technologies. In this survey, three types of optimization techniques are examined: Lyapunov, Meta-heuristic, and Convex. The objective functions, application domains, offloading methods, evaluation techniques, merits, and demerits of the existing algorithms are summarized. The proposed survey undergoes four main CO processes: task scheduling, priority-aware task scheduling, load balancing, and edge server selection. In addition, this research provides a comparative analysis of the offloading techniques. Subsequently, unresolved problems and upcoming unfocused or inadequately covered research tasks are explored. © The Institution of Engineers (India) 2025.

关键词： Computation offloading

来源：评论

学校读者我要写书评

暂无评论

Isomorphic Classification of First Degree Cellular Automata

引用

Complex Systems 2025年第1期34卷 59-90页

作者： Vikrant, Vicky Bhattacharjee, Kamalika Mukherjee, Sukanya Department of Computer Science and Engineering National Institute of Technology Tamil Nadu Tiruchirappalli620015 India Department of Computer Science and Engineering Institute of Engineering and Management West Bengal Kolkata700091 India

This paper explores the concept of isomorphism in cellular automata (CAs), focusing on identifying and understanding isomorphic relationships between distinct CAs. A cellular automaton (CA) is said to be isomorphic to another CA if the configuration transition diagrams of the two CAs are identical. To illustrate how isomorphism entails the preservation of structural properties while allowing for variations in state representations, we exhaustively explore the isomorphic reversible rules in the first degree CAs (FDCAs), that is, the simple three-neighborhood CAs represented by eight parameters. The experimental study includes reversible FDCAs for states two to 10 with cell lengths of four to seven under the null boundary condition and states two to seven with cell lengths of four to seven under the periodic boundary condition. Given a CA, we also provide two synthesis methods for identifying the set of CAs that are isomorphic to it. The first method operates by permuting the states of a group of CA cell locations, whereas the second method involves exchanging cells at various points. We demonstrate that, depending on the range of exchanged cell positions, the isomorphic CAs produced in this way may have a bigger neighborhood than the input CA. Finally, we provide an exhaustive classification of the elementary CAs (ECAs) using the first synthesis method and compare it with existing results. © 2025, Complex Systems Publications, Inc. All rights reserved.

关键词： classification cycle structure FDCA first degree cellular automaton parameters reversibility synthesis schemes

来源：评论

学校读者我要写书评

暂无评论

A Novel CAPTCHA Recognition System Based on Refined Visual Attention

引用

computers, Materials & Continua 2025年第4期83卷 115-136页

作者： Zaid Derea Beiji Zou Xiaoyan Kui Monir Abdullah Alaa Thobhani Amr Abdussalam School of Computer Science and Engineering Central South UniversityChangsha410083China College of Computer Science and Information Technology Wasit UniversityWasit52001Iraq Department of Computer Science and Artificial Intelligence College of Computing and Information TechnologyUniversity of BishaBisha67714Saudi Arabia Electronic Engineering and Information Science Department University of Science and Technology of ChinaHefei230026China

Improving website security to prevent malicious online activities is crucial,and CAPTCHA(Completely Automated Public Turing test to tell computers and Humans Apart)has emerged as a key strategy for distinguishing human users from automated ***-based CAPTCHAs,designed to be easily decipherable by humans yet challenging for machines,are a common form of this ***,advancements in deep learning have facilitated the creation of models adept at recognizing these text-based CAPTCHAs with surprising *** our comprehensive investigation into CAPTCHA recognition,we have tailored the renowned UpDown image captioning model specifically for this *** approach innovatively combines an encoder to extract both global and local features,significantly boosting the model’s capability to identify complex details within CAPTCHA *** the decoding phase,we have adopted a refined attention mechanism,integrating enhanced visual attention with dual layers of Long Short-Term Memory(LSTM)networks to elevate CAPTCHA recognition *** rigorous testing across four varied datasets,including those from Weibo,BoC,Gregwar,and Captcha 0.3,demonstrates the versatility and effectiveness of our *** results not only highlight the efficiency of our approach but also offer profound insights into its applicability across different CAPTCHA types,contributing to a deeper understanding of CAPTCHA recognition technology.

关键词： Text-based CAPTCHA recognition refined visual attention web security computer vision

来源：评论

学校读者我要写书评

暂无评论

Performance optimization in ddos prediction with ensemble based approach

引用

Multimedia Tools and Applications 2025年第12期84卷 10631-10644页

作者： Dogra, Amit Taqdir Department of Computer Science and Engineering SoET BGSB University Rajouri India Department of Computer Science Engineering and Technology Guru Nanak Dev University Regional Campus Gurdaspur India

Distributed Denial of Service (DDoS) attacks pose a significant threat to network infrastructures, leading to service disruptions and potential financial losses. In this study, we propose an ensemble-based approach for DDoS attack detection, leveraging the strengths of three different classifiers: Adaboost, K-Nearest Neighbors (KNN), and Random Forest. We apply data normalization during pre-processing, utilize a Multi-Layer Perceptron (MLP) for feature extraction, and combine the classifiers using an ensemble approach. The performance of each individual classifier and the ensemble is extensively evaluated, and the results demonstrate the effectiveness of the proposed ensemble in accurately identifying and mitigating DDoS attacks. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2024.

关键词： Denial-of-service attack

来源：评论

学校读者我要写书评

暂无评论

Predicting early ASD traits of adults and toddlers using machine learning and deep learning with explainable AI and optimization

引用

Neural Computing and Applications 2025年 1-28页

作者： Rahman, Md. Ashiqur Hossain, Md. Mamun Singh, Sondip Poul Sharmin, Nusrat Dhaka1216 Bangladesh Department of Computer Science and Engineering Military Institute of Science and Technology Dhaka Bangladesh

Autism spectrum disorder (ASD) is a complex neurodevelopmental condition characterized by challenges in social interaction, communication difficulties, repetitive behaviors, and a range of strengths and differences in cognitive abilities. Early ASD diagnosis using machine learning and deep learning techniques is crucial for preventing its severity and long-term effects. The articles published in this area have only applied different machine learning algorithms, and a notable gap observed is the absence of an in-depth analysis in terms of hyperparameter tuning and the type of dataset used in this context. This study investigated predictive modeling for ASD traits by leveraging two distinct datasets: (i) a raw CSV dataset with tabular data and (ii) an image dataset with facial expression. This study aims to conduct an in-depth analysis of ASD trait prediction in adults and toddlers by doing hyper optimized and interpreting the result through explainable AI. In the CSV dataset, a comprehensive exploration of machine learning and deep learning algorithms, including decision trees, Naive Bayes, random forests, support vector machines (SVM), k-nearest neighbors (KNN), logistic regression, XGBoost, and ANN, was conducted. XGBoost emerged as the most effective machine learning algorithm, achieving an accuracy of 96.13%. The deep learning ANN model outperformed the traditional machine learning algorithms with an accuracy of 99%. Additionally, an ensemble model combining a decision tree, random forest, SVM, KNN, and logistic regression demonstrated superior performance, yielding an accuracy of 96.67%. The XGBoost model, utilized in hyperparameter optimization for CSV data, exhibited a substantial accuracy increase, reaching 98%. For the image dataset, advanced deep learning models, such as ResNet50, VGG16, Boosting, and Bagging, were employed. The bagging model outperformed the others, achieving an impressive accuracy of 99%. Subsequent hyperparameter optimization was conduct

关键词： Random forests

来源：评论

学校读者我要写书评

暂无评论

Hybrid RMDL-CNN for speech recognition from unclear speech signal

引用

International Journal of Speech technology 2025年第1期28卷 195-217页

作者： Bhargava, Raja Arivazhagan, N. Babu, Kunchala Suresh Research Scholar Department of Computer Science and Engineering SRM Institute of Science and Technology Kattankulathur Chennai India Department of Computational Intelligence SRM Institute of Science and Technology Kattankulathur Chennai India Department of Computer Science and Engineering Potti Sriramulu Chaluvadi Mallikarjuna Rao college of Engineering and Technology Andhra Pradesh Vijayawada India

ASR is an effectual approach, which converts human speech into computer actions or text format. It involves extracting and determining the noise feature, the audio model, and the language model. The extraction and determination of the noise feature is a crucial aspect of speech recognition, serving as both a process of information compression and signal deconvolution. ASR schemes are mostly employed in smart homes, smart appliances, and biometric schemes. Yet, traditional approaches offer very low performance because of a noisy environment. Moreover, local differences and accents negatively influence the ASR scheme execution during the conversion of the speech signals. This paper introduces a hybrid RMDL-CNN method to address these challenges. At first, the input of unclear speech is carried out by the dataset. Then, signal pre-processing is done by employing a Gaussian filter. After that, voice enhancement is accomplished by employing nonlinear spectral subtraction. Later, the speech word is segmented from the enhanced output based on the Attentional Encoder-Decoder approach and finally, the speech is recognized using the proposed RMDL-CNN. The RMDL-CNN method is devised by the combination of RMDL and CNN. Furthermore, the established RMDL-CNN is accessed for its efficiency based on several values of k-group value, as well as learning data. In addition, the introduced RMDL-CNN approach for speech recognition achieved better accuracy, PPV, as well as NPV of 0.909, 0.947, and 0.917 for dataset 1. Moreover, the RMDL-CNN has achieved the highest accuracy of 0.909, PPV of 0.926 and NPV of 0.888 for dataset 2. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2025.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Weighted rank aggregation based on ranker accuracies for feature selection

引用

Soft Computing 2025年第4期29卷 1981-2001页

作者： Abdolrazzagh-Nezhad, Majid Kherad, Mahdi Faculty of Engineering Department of Computer Engineering Bozorgmehr University of Qaenat Qaen Iran Department of Computer Science Faculty of Computer and Industrial Engineering Birjand University of Technology Birjand Iran Faculty of Engineering Department of Computer Engineering University of Qom Qom Iran

Rank aggregation is the combination of several ranked lists from a set of candidates to achieve a better ranking by combining information from different sources. In feature selection problem, due to the heterogeneity of methods, there are some base rankers (Filter-based methods) that are of diverse quality and usually the ground truth of ratings is not available. Existing rank aggregation methods that take the diverse quality of base rankers into account do not have any explicit approach for appropriate weighting, require prior assumptions, and suffers from high computational complexity. In this paper, to overcome these challenges, an efficient unsupervised method is introduced for estimating the base rankers’ qualities and aggregating the rankers based on the estimated weights. We first compute the ratio of disagreement between base rankers in ordering different element pairs and then estimate the accuracies in a way that to minimize the discrepancy between these computed ratios and their analytical counterparts. We use the weighted majority voting method for obtaining the aggregated results. To resolve the probable inconsistencies in the final aggregation, the result is formed as a graph, and a greedy algorithm is used to find an acyclic subgraph with the highest weigh. To demonstrate the performance of the proposed method, nine standard UCI datasets are used. The obtained results by the proposed method have higher values of classifier measures than the existing baseline Feature Selection methods and rank aggregation-based multi-filter methods in the most datasets. The experiments show that rank aggregation-based Feature Selection methods outperform individual methods. The proposed method also shows the weight of each Filter-based Feature Selection method, in which the MRMR method has a higher weight than other methods. © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2025.

关键词： Wiener filtering

来源：评论

学校读者我要写书评

暂无评论

ElHBiAt: Electra pre-training network Hybrid of BiLSTM, the Attention layer to Aspect-Based Sentiment Analysis

引用

IEEE Access 2025年 13卷 88342-88370页

作者： Ghanee, Amin Jahanbin, Kia Yamchi, Abolfazl Razzaghi Shahid Beheshti University Department of Computer Engineering Tehran Iran Yazd University Department of Computer Engineering Yazd Iran Shiraz University Department of Computer Science Engineering and Information Technology Shiraz Iran

Aspect-based sentiment analysis is one of the famous and practical subjects in natural language processing. Traditional sentiment analysis assigns a polarity to the whole text or document and does not consider the aspects of the text, so it cannot correctly show the author’s opinion. Extracting aspects and determining their polarity has many challenges, such as the lack of rich datasets, negative transfer in cross-domain datasets, and fine-tuning of deep models of the ratio of texts with different contexts. In this research, a model has been proposed for aspect-based sentiment analysis with different contexts. This model uses the Electra as the backbone of the architecture, which is trained with different cross-domain datasets to transfer knowledge to the target domain. Also, in the proposed model of ElHBiAt, the resampling method is used to deal with imbalanced datasets, which causes the neural model to find less orientation than the polarities in the majority. In this model, two layers of attention are used to extract the essential aspects of the text, and the deep neural network BiLSMT is used to process the text in a two-way. The results show that the ElHBiAt model has been able to improve the efficiency by an average of 5.78% in the SemEval dataset, by 2% in the SEntFiN dataset, and by an average of 4.3% in the Drug dataset in terms of the F1-score. © 2013 IEEE.

关键词： Emotion Recognition

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：