检索结果-内蒙古大学图书馆

21st Pacific Rim International Conference on Artificial Intelligence, PRICAI 2024

作者： Tan, Yajing Huang, Yuwei Duan, Qiqi Yang, Yijun Shi, Yuhui Department of Computer Science and Engineering Southern University of Science and Technology Shenzhen China

ISBN: (纸本)9789819601219

Very recently, a memory-efficient version (called MeZO) of simultaneous perturbation stochastic approximation (SPSA), one well-established zeroth-order optimizer from the automatic control community, has shown competitive fine-tuning performance on large language models (LLMs) with billions of weights under the limited memory of a single GPU, which often hinders the direct usage of the standard backpropagation (BP) update. Furthermore, population-based evolutionary methods (such as evolution strategies from OpenAI and population-based training from DeepMind) have shown a promising direction to improving and scaling up various large-scale AI systems via efficiently utilizing increasingly available distributed computing resources, owing to their relatively simple yet efficient data and control flows. In this paper, we propose a population-based SPSA for fine-tuning LLMs with non-differentiable objective functions. Specifically, our proposed method combines the desirable exploration ability of distributed population-based evolutionary methods with the powerful exploitation ability of SPSA in the modern distributed computing environment wherein the efficient implementation of BP is still challenging because of its complex dependencies in forward-pass and backward-pass information flows. Simulation experiments on a set of challenging (non-differentiable) natural language process tasks empirically validate the potential of our method on fine-tuning even in a relatively small-scale distributed computing scenario. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Stochastic systems

来源：评论

学校读者我要写书评

暂无评论

A QoS-aware routing approach for Internet of Things-enabled wireless sensor networks in smart cities

引用

Multimedia Tools and Applications 2025年第17期84卷 17951-17977页

作者： Karunkuzhali, D. Meenakshi, B. Lingam, Keerthi Department of Information Technology Panimalar Engineering College Chennai India Department of Electrical and Electronics Engineering Sri Sairam Engineering College Chennai India Department of Computer Science and Engineering Gitam University Visakhapatnam India

In the Internet of Things (IoT), optimizing machine performance through data analysis and improved connectivity is pivotal. Addressing the growing need for environmentally friendly IoT solutions, we focus on "green IoT." Quality in smart cities heavily relies on efficient data gathering and communication. In this study, we propose a novel software architecture tailored for data collection and communication in IoT-enabled smart applications, emphasizing sustainability from the outset. Specifically, for smart cities utilizing IoT-enabled wireless sensor networks, we introduce a Quality of Service (QoS)-aware routing strategy. In the data collection phase, we employ Chaotic Bird Swarm Optimization (CBSO) for creating IoT sensor clusters, accompanied by Improved Differential Search (IDS) to assess the reliability of individual sensor nodes, ultimately designating the most trusted node as the Cluster Head (CH). In the data transport phase, we implement lightweight signcryption to enhance the security of IoT sensor data. Furthermore, we employ an Optimal Data Routing (ODM) approach to compute the most efficient data transmission paths between source and destination nodes within the IoT network. Finally, the performance of our QoS-aware routing strategy is evaluated through network simulations using NS2 and compared against existing approaches, offering valuable insights into its effectiveness and suitability for smart city IoT applications. The main findings of the study indicate that the proposed QoS-aware routing strategy (QOS-AWARE ROUTING) consistently outperforms existing routing algorithms in terms of energy efficiency, network longevity, and latency, both in scenarios with varying sensor node density and over multiple cycles. Additionally, it demonstrates significant advantages, such as up to 90.8% less energy consumption, 66.8% longer network lifespan, and 80.1% reduced latency when compared to the benchmarked methods. © The Author(s), under exclusive licence to S

关键词： Internet of things

来源：评论

学校读者我要写书评

暂无评论

Diluie: constructing diverse demonstrations of in-context learning with large language model for unified information extraction

引用

Neural Computing and Applications 2024年第22期36卷 13491-13512页

作者： Guo, Qian Guo, Yi Zhao, Jin Department of Computer Science and Engineering East China University of Science and Technology Shanghai200237 China Shanghai Key Laboratory of Data Science School of Computer Science Fudan University Shanghai200433 China

Large language models (LLMs) have demonstrated promising in-context learning capabilities, especially with instructive prompts. However, recent studies have shown that existing large models still face challenges in specific information extraction (IE) tasks. Moreover, it could have more effectively utilized various prompts such as instruction tuning, diverse demonstrations of in-context learning, and long-range token sequences for assisting language modeling in understanding context. In this study, we propose DILUIE, a unified information extraction framework based on in-context learning with diverse demonstration examples. DILUIE is encoded with an EVA attention mechanism and incremental encoding technology. Based on the constructed diverse demonstrations, we expand the size of instances efficiently in both instruction tuning and in-context learning to gain insights into the potential benefits of utilizing diverse information extraction datasets. To deepen the understanding of context, we further design three auxiliary tasks to assist in aligning contextual semantics. Experimental results demonstrate that DILUIE achieves 2.23 and 2.53% improvements in terms of Micor-/Macor-F1 on average relative to the current state-of-the-art baseline, which also significantly outperforms the GPT-3.5-turbo in zero-shot settings, and the average token length of achieving the best performance over tasks is around 15k. Furthermore, we observe that in-context learning shows enhanced performance when provided with more demonstrations during multiple-shot instruction tuning (8 k). Additionally, increasing the length of instructions (10 k) can result in a more substantial improvement in the upper limits of scaling for in-context learning. Code is available on https://***/Phevos75/DILUIE. © The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2024.

关键词： Demonstrations

来源：评论

学校读者我要写书评

暂无评论

Enhancing Occluded Face Recognition Through Device-Edge Collaboration and Cross-Domain Feature Fusion with Heatmaps 3

Enhancing Occluded Face Recognition Through Device-Edge Coll...

引用

3rd International Conference on Intelligent Data Communication Technologies and Internet of Things, IDCIoT 2025

作者： Prasanth, Ashwin Annie Micheal, A. Department of Computer Science & Engineering Sathyabama Institute of Science & Technology Chennai India

ISBN: (纸本)9798331527549

Face recognition under occlusion presents a persistent challenge in computer vision, primarily due to difficulties in capturing and effectively integrating visible and obscured facial features. This paper introduces a novel device-edge collaborative framework that advances occluded face recognition through an innovative cross-domain feature fusion technique. Our method employs a software-generated, occlusion-aware heatmap created from standard visible-light images, which emphasizes key facial landmarks even when regions of the face are partially covered. This heatmap provides a complementary data layer that enhances traditional visible facial features, allowing the cross-domain feature fusion process to synthesize a more comprehensive set of features for reliable recognition under occlusion. The combined feature set leverages both visible and simulated heatmap information to improve model robustness, even in complex occlusion scenarios. Additionally, a delay-optimized task scheduling method is applied on the edge, ensuring efficient resource allocation, faster processing, and minimal latency. Although this project is ongoing, preliminary evaluations indicate the potential effectiveness of this approach, with expected outcomes to include high recognition accuracy and reduced latency. This framework provides a scalable solution for occluded face recognition by integrating complementary data sources without requiring specialized hardware. © 2025 IEEE.

关键词： Face recognition

来源：评论

学校读者我要写书评

暂无评论

ComStreamClust: a Communicative Multi-Agent Approach to Text Clustering in Streaming Data

引用

Annals of Data science 2023年第6期10卷 1583-1605页

作者： Najafi, Ali Gholipour-Shilabin, Araz Dehkharghani, Rahim Mohammadpur-Fard, Ali Asgari-Chenaghlu, Meysam Department of Computer Science and Engineering Sabanci University Istanbul Turkey Department of Computer Engineering University of Tabriz Tabriz Iran Computer Engineering Department Isik University Istanbul Turkey Department of Computer Engineering Sharif University of Technology Tehran Iran

Topic detection is the task of determining and tracking hot topics in social media. Twitter is arguably the most popular platform for people to share their ideas with others about different issues. One such prevalent issue is the COVID-19 pandemic. Detecting and tracking topics on these kinds of issues would help governments and healthcare companies deal with this phenomenon. In this paper, we propose a novel, multi-agent, communicative clustering approach, so-called ComStreamClust for clustering sub-topics inside a broader topic, e.g., the COVID-19 and the FA CUP. The proposed approach is parallelizable, and can simultaneously handle several data-point. The LaBSE sentence embedding is used to measure the semantic similarity between two tweets. ComStreamClust has been evaluated by several metrics such as keyword precision, keyword recall, and topic recall. Based on topic recall on different number of keywords, ComStreamClust obtains superior results when compared to the existing methods. © 2022, The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.

关键词： COVID-19

来源：评论

学校读者我要写书评

暂无评论

Feature engineering Techniques for Stegware Analysis: An Extensive Survey 1

引用

1st International Conference on Intelligent Computing for Sustainable Development, ICICSD 2023

作者： Anitha, M. Azhagiri, M. Department of Computer Science and Engineering SRM Institute of Science and Technology Ramapuram Chennai India Department of Computer Science and Engineering SRM Institute of Science and Technology Ramapuram Chennai India

ISBN: (数字)9783031612985

ISBN: (纸本)9783031612978

The art of hiding secret text within an innocuous cover medium is steganography. Steganalysis is the counterpart of steganography which focuses on the detection and extraction of the secret text from the medium. Feature engineering is the crucial field in Stegware Analysis which intends to identify more specific features, focusing on the accuracy and efficiency. Feature engineering is a process in Machine learning where the features of any dataset are selected and extracted for further use. Feature engineering is the process of extracting, transforming and selecting the most relevant features form the data that aids in discriminating between the stego and cover image. This is because, most of the time, the data will be in a raw format. Any ML model needs the data to be pre-processed and kept ready to train the model. Thus, from the pool of raw data, the required data needs to be selected and can be used in training the model. Further, the data at point needs to be extracted to get the precise data. The scope of the work is to identify the various feature engineering techniques available in practice and efficiently use them to achieve high accuracy and precision in the system. The survey focuses on the several feature selection and extraction techniques like filter method, wrapper method and embedded methods. Correlation being one of the feature selection methods is focused;while statistical moments computes the mean, variance and skewness of the feature. The extraction method holds the Computation of Invariants and other such. Comparative study is made on both the methods to understand the concepts with ease. The work starts by taking a sample from the dataset and few feature extraction techniques are applied on the same. Then the original image is compared with the extracted images with the view of histogram. The paper gives valuable insights into the effectiveness of different feature engineering techniques using the dataset and underscores the importance of featu

关键词： Steganography

来源：评论

学校读者我要写书评

暂无评论

Automated Brain Tumor Detection and Classification in MRI Scans using a Deep Convolutional Neural Network and U-Net Architecture 6

Automated Brain Tumor Detection and Classification in MRI Sc...

引用

6th International Conference on Mobile Computing and Sustainable Informatics, ICMCSI 2025

作者： Devadas, Prathima Mathivanan, G. Department of Computer science and Engineering Sathyabama Institute of Science and Technology Chennai India

ISBN: (纸本)9798331522667

Brain tumor identification and categorization serve a precarious role in early diagnosis and treatment planning. This study suggests a new deep learning (DL)-based structure for the automatic detection and categorization of brain cancers using medical imaging technologies. The framework uses a deep convolutional neural network (DCNN) with an exclusive segmentation technique, notably the U-Net architecture, to discrete tumor areas in magnetic resonance imaging (MRI) scans. The DCNN model, which contains multiple convolutional layers, is used for feature extraction, followed by fully connected layers for tumor classification. To increase segmentation accuracy, the U-Net model integrates both encoder-decoder structures and skip connections to capture fine-grained spatial details. The suggested algorithm is trained and evaluated on a large dataset of brain MRI images, achieving high accuracy, sensitivity, and specificity for both tumor detection and classification. Experimental results show that the model outclasses standard methods, providing a dependable and operative solution for the early detection and diagnosis of brain cancers, potentially aiding clinical decision-making processes. © 2025 IEEE.

关键词： Magnetic resonance imaging

来源：评论

学校读者我要写书评

暂无评论

Cloud Computing Network in Remote Sensing-Based Climate Detection Using Machine Learning Algorithms

引用

Remote Sensing in Earth Systems sciences 2025年第2期8卷 400-408页

作者： Srinivas, Jhade Raju, Ch VV Narasimha Sasikala, C. Bhaskar, Parumanchala Panyala, Amarendra Reddy Priya, Divya Department of Computer Science and Engineering - Data Science K G Reddy College of Engineering and Technology Moinabad Telangana Hyderabad India Information Technology Marri Laxman Reddy Institute of Technology and Management Dundigal Telangana Hyderabad 500043 India Dept of CSE Srinivasa Ramanujan Institute of Technology (Autonomous) Andhra Pradesh Ananthapuramu India Department of Computer Science and Engineering Santhiram Engineering College Nandyal India Department of Computer Science and Engineering (Data Science) Malla Reddy University Telangana Hyderabad India Department of Computer Science and Engineering MLR Institute of Technology Hyderabad India

Effective monitoring of the environment over a large area will require mobilization of a considerable amount of information. Otherwise, the use of traditional methods will prove to be costly and would take up so much time. Also, assessing the land cover changes between two times is one of the other methods which can be adopted to measure the environmental sustainability. Using ML techniques for predictive modeling in remote sensing has emerged as a useful approach in solving some of the environmental and climatic challenges. This research proposes a novel method in remote sensing image-based climate detection utilizing cloud computing and machine learning algorithms. Here, the cloud computing module is used in the collection of remote sensing images for the detection of climate changes. Then, the collected image is processed for noise removal as well as normalization. The processed image is segmented, and features are extracted utilizing a fuzzy K-clustering-based contour Gaussian model with graph cut Boltzmann CNN. Experimental analysis is carried out for various remote sensing-based climate change datasets in terms of training accuracy, random precision, AUC, recall, and F-measure. The proposed technique obtained 97% of training accuracy, 95% of random precision, 96% of recall, 93% of AUC, and 94% of F-measure. © The Author(s), under exclusive licence to Springer Nature Switzerland AG 2025.

关键词： Climate detection Cloud computing Contour Gaussian model Machine learning algorithms Remote sensing images

来源：评论

学校读者我要写书评

暂无评论

Elevating IoT healthcare security using ProSRN and ICOM methodologies for effective threat management

引用

International Journal of Information technology (Singapore) 2025年 1-19页

作者： Sowjanya, Yemineni Gopalakrishnan, S. Kumar, R. Dinesh Department of Computer Science and Engineering Hindustan Institute of Technology and Science Tamil Nadu Chennai 603103 India Department of Information Technology Hindustan Institute of Technology and Science Tamil Nadu Chennai 603103 India Department of Computer Science and Engineering Bhoj Reddy Engineering College for Women Telangana Hyderabad 500059 India

IoT healthcare security is increasingly important, as the interconnectedness of medical devices in itself introduces a major vulnerability, given the impact on patient safety and data integrity. Past works in this domain have been afflicted by persistent issues of high false-positive rates, resource utilization inefficiency, and failure to adapt to rapidly changing threat spaces. The paper presents a novel framework that incorporates ProSRN, which provides the robustness in threat detection, and ICOM for dynamic weight estimation. ProSRN adopts a residual-connected stacking approach to reach high accuracy of detection while reducing the complexity in training and enhancing responsiveness in real time. It is further complemented by ICOM, which will then dynamically optimize network parameters to yield an adaptive IoT health security model with the capability of addressing evolving threats. Together, ProSRN and ICOM advance both detection and optimization significantly, enhancing the adaptiveness, efficiency, and security in IoT healthcare systems. This creates a concrete basis for developing resilient frameworks that can protect sensitive healthcare data against emerging threats. © Bharati Vidyapeeth's Institute of computer Applications and Management 2025.

关键词： And deep learning Attack detection Classification Healthcare systems Internet of things (IoT) Optimization Security

来源：评论

学校读者我要写书评

暂无评论

Kookaburra Optimization Algorithm and Feature Aggregated Region based Differential network for Colon Cancer Detection from Histopathological Images 8

Kookaburra Optimization Algorithm and Feature Aggregated Reg...

引用

8th International Conference on I-SMAC (IoT in Social, Mobile, Analytics and Cloud), I-SMAC 2024

作者： Gowthamy, J. Ramesh, S.S. Subashka Rm Institute of Science and Technology Amapuram Department of Computer Science and Engineering Chennai India Srm Institute of Science and Technology Ramapuram Department of Computer Science and Engineering Chennai India

ISBN: (纸本)9798350376425

Colon cancer is a significant health concern, ranking as the third leading cause of cancer-related deaths in men and the second in women. Early detection is crucial in reducing the incidence and mortality rates associated with this malignancy. This paper introduces a novel Feature Aggregated Region-based Differential Network (FA-RDNet) designed for the accurate classification of colon cancer from histopathological images. Our approach leverages Multiple Instance Learning (MIL) to effectively handle large datasets structured as bags of instances, addressing the limitations of conventional MIL techniques. A key innovation of our work is the integration of an adaptive feature enhancement module that bilaterally updates multi-level deep features, allowing for efficient extraction of both low-level details and high-level semantic information. Furthermore, we introduce the Kookaburra Optimization Algorithm (KOA) to fine-tune the parameters of FA-RDNet, utilizing a unique two-phase mathematical modeling process inspired by the hunting behaviors of kookaburras. This enhances the model's optimization and classification accuracy. The effectiveness of our proposed system is validated using the publicly available LC25000 dataset, where it achieved a remarkable accuracy of 94.42%, outperforming existing state-of-the-art methods in colon cancer detection. The high computational efficiency and accuracy of FA-RDNet highlight its potential for practical applications in clinical settings, marking a significant advancement in the field of medical image analysis. © 2024 IEEE.

关键词： Lung cancer

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：