检索结果-内蒙古大学图书馆

2024 International Conference on Computational Intelligence and Computing Applications, ICCICA 2024

作者： Modi, Soham Kumar Chahat Yadav, Anup Lal Rana, Pratham Singh Kalpana Chandigarh University Department Of Computer Science And Engineering Punjab India

ISBN: (纸本)9798350306446

The significance of soil in sustaining life on Earth cannot be overstated, particularly its role in crop production and environmental sustainability. This paper presents a analysis of various machine-learning models employed in soil classification and their implications for crop productivity. Soil texture and nutrient composition are pivotal factors influencing crop growth and sustainability. Leveraging machine learning techniques, such as classification algorithms and computer vision, this study evaluates the effectiveness of Decision Trees (DT), k nearest Neighbors (k-NN), Artificial Neural Networks (ANN), and Support Vector Machines (SVM) in automating the soil classification process. These models utilize detectable features including Soil characteristics like moisture, nutrients, structure, quality, pH, and texture to classify soil types accurately. Additionally, techniques like interpolation of soil particle size fractions using transformed data are explored. By examining the performance and efficiency of these models, this research aims to contribute insights into optimizing soil management practices for enhanced crop production and environmental sustainability. © 2024 IEEE.

关键词： Neural networks

来源：评论

学校读者我要写书评

暂无评论

IT-Tuning: Parameter Efficient Information Token Tuning for Language Model 9

IT-Tuning: Parameter Efficient Information Token Tuning for ...

引用

9th Workshop on Representation Learning for NLP, RepL4NLP 2024 at ACL 2024

作者： Kim, Jungu Kim, Hyeoncheol Department of of Computer Science and Engineering Korea University Korea Republic of

ISBN: (纸本)9798891761575

Recently, language models have demonstrated exceptional performance compared to their predecessors. In this context, attention mechanisms and pre-training significantly contribute to the enhanced performance of modern language models. Additionally, a continuously increasing number of parameters plays a crucial role in these advancements. However, an increase in the number of parameters significantly increases the GPU memory and training time required during fine-tuning of language models, this makes fine-tuning infeasible in environments with limited computing resources. Furthermore, after fine-tuning, the storage space required for deployment increases proportionally with the number of tasks, making it challenging to deploy devices with limited storage capacities. In this study, we propose IT-Tuning, a Parameter Efficient Fine-Tuning method that introduces a new concept called information tokens to address these issues. © 2024 Association for Computational Linguistics.

关键词： Computational linguistics

来源：评论

学校读者我要写书评

暂无评论

Optimizing Block Propagation in Bitcoin Network with Region-based Neighbor Selection Using Reinforcement Learning 6

Optimizing Block Propagation in Bitcoin Network with Region-...

引用

6th IEEE International Conference on Blockchain and Cryptocurrency, ICBC 2024

作者： Choi, Wonseok Jeong, Euidong Woo, Jongsoo Hong, James Won-Ki Postech Department of Computer Science and Engineering Pohang Korea Republic of

ISBN: (纸本)9798350316742

Bitcoin proved the potential of blockchain technology through decentralized, transparent, and immutable transactions. However, there are still challenges for fast and stable transitions. Optimizing block propagation times within the network is one of them. Prolonged propagation times can restrict efficiency, scalability, and security. This paper presents a novel approach to reducing block propagation time through leveraging reinforce-ment learning (RL) for the node's neighbor selection strategies. We implemented a Deep Q-Network (DQN) model in minimizing block receive times at each node, thereby impacting overall block propagation time. We used a model that defines node states based on latencies of outbound connections, which present the node's region. By evaluating this model through simulations using SimBlock, a robust Bitcoin network simulator, we observed a significant reduction in block propagation time - approximately 30% for smaller networks and 20% for larger ones. Our analysis extended to node connections generated by our model and comparative evaluation against existing methodologies. © 2024 IEEE.

关键词： Deep reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

Tomato Leaf Disease Detection and Classification by Using Novel CNN Model 5th

Tomato Leaf Disease Detection and Classification by Using No...

引用

5th International Conference on computer and Communication Technologies, IC3T 2023

作者： Rani, D. Sandhya Shyamala, K. Department of Computer Science and Engineering Osmania University Hyderabad India

ISBN: (纸本)9789819997060

The recognition and early prevention of leaf diseases on time is very essential for improving the crop production. Different deep learning models are proposed by the researchers for identification of diseases in plants. A novel CNN model is implemented in this paper to identify and diagnose plant diseases. The standard CNN model requires a huge no. of parameters, cost of computation is high and also training time is more. In the place of the standard convolution depthwise separable convolutions are used in order to minimize the no. of parameters and computation cost. The implemented CNN architecture attained better classification accuracy of 99.05% when compared with other models like standard inception V3 98.45%, ResNet101V2 98.52%, EfficientNetB0 99.34%, ResNet 50V2 99.46%, and DenseNet121 99.60%. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024.

关键词： Convolution

来源：评论

学校读者我要写书评

暂无评论

Twitter Based Emotion Recognition Using Bi-LSTM 2

Twitter Based Emotion Recognition Using Bi-LSTM

引用

2nd IEEE International Conference on Trends in Quantum Computing and Emerging Business Technologies, TQCEBT 2024

作者： Anbazhagan, K. Kurlekar, Supriya Brindha, T.V. Sudhish Reddy, D. Saveetha School of Engineering Department of Computer Science and Engineering Chennai India Department of Electronics and Telecommunication Engineering JSPM Narhe Technical Campus Pune India Sri Venkateshwara College of Engineering and Technology Department of Computer Science and Engineering Tiruvallur India Rmd Engineering College Department of Computer Science and Engineering Tiruvallur India

ISBN: (纸本)9798350384277

Twitter and other social media platforms have evolved into crucial sources of information and communication for billions of people across the world. The massive amount of data collected on these platforms enables insights into people's emotions and feelings. Emotion identification, in particular, has gotten a lot of attention because of its potential uses in industries like marketing, healthcare, and social study. "Emotional states of individuals, also known as moods, are central to expressing thoughts, ideas, and opinions, which influence attitudes and behavior."In this study, we present a method for identifying the emotion or mood of a tweet and classifying it into the appropriate emotional category. The work mentioned in this paper aims to identify and analyze the mood and emotion displayed by people in their text tweets to offer ideas. We gathered tweets and replies on a particular set of themes to create a dataset with text, user, emotion, sentiment, and other data. We used the dataset to compute user impact ratings based on various user and tweet-related factors to analyze feelings and emotions in tweets and their replies. Our model used a brand-new bidirectional long short-term memory (BiLSTM) and previously learned word embedding vectors. Finally, using the latest data, we made generalized and specific consumer suggestions based on Twitter's conduct. According to the experimental findings, the suggested model has 95.56% and 99.16% prediction accuracy. Modern performance improvement was achieved for the model. © 2024 IEEE.

关键词： Tweets

来源：评论

学校读者我要写书评

暂无评论

Effective Malware Classification using Fine-tuned CNN Architecture: An Image-Based Approach 2

Effective Malware Classification using Fine-tuned CNN Archit...

引用

2nd IEEE International Conference on Networking and Communications, ICNWC 2024

作者： Sasikala, L. Shanmuganathan, C. SRM Institute of Science and Technology Department of Computer Science and Engineering Ramapuram Chennai India

ISBN: (纸本)9798350365269

Malware detection and classification are crucial in cybersecurity to protect systems and networks against malicious attacks. This study introduces a proficient image-based method employing convolutional neural network (CNN) architecture. It aims to enhance malware classification by addressing the challenge of detecting polymorphic or obfuscated malware variants, which can be difficult with conventional techniques reliant on static or dynamic analysis of executable files. Our proposed model tackles this constraint by representing malware samples as grayscale and RGB images, capturing both structural and visual aspects. To improve classification accuracy, we fine-tuned a pre-trained CNN model on a Malign dataset of 25 families and 9342 malware samples, leveraging transfer learning. Our experimental results show that our method accurately classifies distinct types of malware with excellent precision and recall rates. Additionally, image-based representation is resistant to evasion strategies used by advanced malware. Our proposed approach demonstrates potential for enhancing malware detection and classification in practical cybersecurity settings © 2024 IEEE.

关键词： Cybersecurity

来源：评论

学校读者我要写书评

暂无评论

Enhancing Deep Learning Model Privacy Against Membership Inference Attacks Using Privacy-Preserving Oversampling

引用

SN computer science 2025年第4期6卷 1-21页

作者： Ghosh, Subhasish Mandal, Amit Kr Cortesi, Agostino Department of Computer Science and Engineering SRM University AP Andhra Pradesh Amaravati 522240 India Department of Computer Science Ca’ Foscari University Via Turino 155 Venice 30170 Italy

The overfitting of deep learning models trained using moderately imbalanced datasets is the main factor in increasing the success rate of membership inference attacks. While many oversampling methods have been designed to minimize the data imbalance, only a few defend the deep neural network models against membership inference attacks. We introduce the privacy preserving synthetic minority oversampling technique (PP-SMOTE), that applies privacy preservation mechanisms during data preprocessing rather than the model training phase. The PP-SMOTE oversampling method adds Laplace noise to generate the synthetic data points of minority classes by considering the L1 sensitivity of the dataset. The PP-SMOTE oversampling method demonstrates lower vulnerability to membership inference attacks than the DNN model trained on datasets oversampled by GAN and SVMSMOTE. The PP-SMOTE oversampling method helps retain more model accuracy and lower membership inference attack accuracy compared to the differential privacy mechanisms such as DP-SGD, and DP-GAN. Experimental results showcase that PP-SMOTE effectively mitigates membership inference attack accuracy to approximately below 0.60 while preserving high model accuracy in terms of AUC score approximately above 0.90. Additionally, the broader confidence score distribution achieved by the PP-SMOTE significantly enhances both model accuracy and mitigation of membership inference attacks (MIA). This is confirmed by the loss-epoch curve which shows stable convergence and minimal overfitting during training. Also, the higher variance in confidence scores complicates efforts of attackers to distinguish training data thereby reducing the risk of MIA. © The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd. 2025.

关键词： Deep neural networks Differential privacy Membership inference attack Oversampling method

来源：评论

学校读者我要写书评

暂无评论

AI-Powered Medical Record Summarization for Diagnosis and Physician Training 3

AI-Powered Medical Record Summarization for Diagnosis and Ph...

引用

3rd International Conference on Applied Artificial Intelligence and Computing, ICAAIC 2024

作者： Singh, Abhishek Rawat, Abhas Khurana, Vrinda Chandel, Jyoti Chandigarh University Department of Computer Science Engineering Mohali India

ISBN: (纸本)9798350375190

In this article, researchers discuss the ways in which computers can assist doctors in diagnosing diseases by summarizing medical records. They also explore the impact of using computers to analyze health records on healthcare and on patients' ability to comprehend their health. The article discusses utilizing internet search data for early disease detection and addresses how searching for health information online may lead to excessive worrying. It also discusses various computer software that assist in diagnosing illnesses, including newer ones such as Isabel. The paper discusses the utilization of sophisticated computer methods such as neural networks to improve comprehension of medical data. It evaluates the ability of these methods to categorize medical text and assign codes for diagnoses. The article also discusses a computer program that accurately diagnoses knee pain and examines the use of deep learning in medical research. The paper also addresses methods to prevent computer programs from becoming too fixated on particular cases and techniques for visualizing data to enhance comprehension. In general, the goal of the paper is to enhance healthcare using computers to aid in more accurate disease diagnosis by doctors and support in medical research. Researchers in this paper explore how artificial intelligence (AI) could transform healthcare practices. They explore how AI can simplify the task of summarizing medical records, helping doctors make faster and more precise diagnoses. Furthermore, the article explores how using electronic health records (EHRs) can transform healthcare delivery and empower patients. © 2024 IEEE.

关键词： Diagnosis

来源：评论

学校读者我要写书评

暂无评论

An Approach for Adaptive Traffic Light Control System in ITS using VANET 2

An Approach for Adaptive Traffic Light Control System in ITS...

引用

2nd International Conference on Emerging Trends in Information Technology and engineering, ic-ETITE 2024

作者： Islam, Md. Mazaherul Rehena, Zeenat Mukherjee, Nandini Jadavpur University Department of Computer Science and Engineering Kolkata India

ISBN: (纸本)9798350328202

Traffic congestion is a common problem in everyday's life as number of vehicles is increasing day by day on the road. So, traditional road traffic control system needs to be updated accordingly. An adaptive traffic light control system can solve the traffic congestion problem efficiently. Vehicular Adhoc Network (VANET) is a common and useful application for designing an adaptive traffic light control system as a part of Intelligent Transportation System (ITS). This paper presents a literature review work on different types of traffic control systems and proposes a method for designing an adaptive traffic light control system using VANET technology. The proposed system ensures that all the vehicles pass through a road intersection point in an urban area smartly and efficiently. It also reduces traffic congestion and solves some other problems related to congestion by improving traffic flow through an intersection point in a city area. © 2024 IEEE.

关键词： Traffic congestion

来源：评论

学校读者我要写书评

暂无评论

SwiftTap: RFID-Driven Automation for Instant Password Entry 8

SwiftTap: RFID-Driven Automation for Instant Password Entry

引用

8th IEEE International Conference on Computational System and Information Technology for Sustainable Solutions, CSITSS 2024

作者： Manhas, Prabal Anandu, U. Hariharan, U. Chandigarh University Department of Computer Science & Engineering Mohali India

ISBN: (纸本)9798331505462

SwiftTap is an innovative approach to user authentication by using RFID technology to replace conventional password-based methods. By employing everyday objects such as keychains and cards as secure substitutes, SwiftTap streamlines the authentication process. By merging progress in hardware and software, SwiftTap improves security and simplifies user access. This research delves into the intricacies of utilizing SwiftTap in different scenarios, such as in business settings and on personal devices. It also examines factors concerning user experience and their potential impacts. SwiftTap signifies a fresh era of effortless authentication, prompting users to embrace a smoother method of entry without the usual login difficulties. In addition to its seamless authentication capabilities, SwiftTap also addresses privacy concerns by employing encryption protocols that safeguard user data during transmission. Its versatility extends to integration with existing systems, promising compatibility across diverse platforms. With SwiftTap, the paradigm shifts towards effortless yet secure authentication is within reach, promising enhanced user. © 2024 IEEE.

关键词： Firmware

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：