检索结果-内蒙古大学图书馆

30th IEEE Real-Time and Embedded Technology and Applications Symposium, RTAS 2024

作者： Kang, Woosung Lee, Jinkyu Lee, Youngmoon Oh, Sangeun Lee, Kilho Chwa, Hoon Sung DGIST Dept. of Electrical Engineering and Computer Science Korea Republic of Dept. of Computer Science and Engineering Korea Republic of Hanyang University Dept. of Robotics Korea Republic of Ajou University Department of Software and Computer Engineering Korea Republic of School of AI Convergence Soongsil University Korea Republic of

ISBN: (纸本)9798350358414

The increasing complexity and memory demands of Deep Neural Networks (DNNs) for real-Time systems pose new significant challenges, one of which is the GPU memory capacity bottleneck, where the limited physical memory inside GPUs impedes the deployment of sophisticated DNN models. This paper presents, to the best of our knowledge, the first study of addressing the GPU memory bottleneck issues, while simultaneously ensuring the timely inference of multiple DNN tasks. We propose RT-Swap, a real-Time memory management framework, that enables transparent and efficient swap scheduling of memory objects, employing the relatively larger CPU memory to extend the available GPU memory capacity, without compromising timing guarantees. We have implemented RT-Swap on top of representative machine-learning frameworks, demonstrating its effectiveness in making significantly more DNN task sets schedulable at least 72% over existing approaches even when the task sets demand up to 96.2% more memory than the GPU's physical capacity. © 2024 IEEE.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

Improved Object Detection Through a Learning-Based Joint-Initiative Supervised Autonomy (JISA) Framework

Improved Object Detection Through a Learning-Based Joint-Ini...

引用

2025 International Conference on Control, Automation, and Instrumentation, IC2AI 2025

作者： Tarabay, Andrea Daher, Naseem American University of Beirut Dept. of Electrical and Computer Engineering Beirut Lebanon American University of Beirut Dept. of Electrical and Computer Engineering Dept. of Mechanical Engineering Beirut Lebanon

ISBN: (纸本)9798331529604

Object detection is faced with accuracy and robustness limitations due to unstructured environments with objects of different sizes and occlusions. Despite advancements in object detection models, these models have not reached a point to possess a comparable level of perception to humans, thus limiting their ability to effectively address challenges in diverse environments. Hence, adopting a human-in-the-loop approach in autonomous systems can handle complex perception scenarios. To tackle these issues, the Joint-Initiative Supervised Autonomy (JISA) framework is used to allow a human supervisor to intervene once the agent is uncertain of its predictions based on its self-confidence (SC) metric, or when the human deems the prediction incorrect based on his/her situation awareness (SA). However, the existing JISA framework formally defines the levels of autonomy between the human and the agent, but does not provide a means for learning from human interventions. In this paper, we present an enhanced version of JISA in object detection applications by implementing a learning scheme that is inspired by the Interactive Corrections and Reinforcement for an Epistemic and Aleatoric uncertainty-aware Teaching (ICREATe) algorithm. To address these challenges, we evaluate the results of the proposed framework and demonstrate that the model effectively learns and improves its performance through human feedback. © 2025 IEEE.

关键词： Self-supervised learning

来源：评论

学校读者我要写书评

暂无评论

UAV-based Firefighting by Multi-agent Reinforcement Learning 13

UAV-based Firefighting by Multi-agent Reinforcement Learning

引用

13th International Conference on computer and Knowledge engineering, ICCKE 2023

作者： Tanha, Reza Shami Hooshmand, Mohsen Afsharchi, Mohsen Dept. Computer Science and It Zanjan Iran University of Zanjan Dept. Electrical and Computer Engineering Zanjan Iran

ISBN: (纸本)9798350330151

Wildfire in the forest leads to a huge amount of financial and human losses. That is it causes damage to the forest and the life of firefighters. To reduce the amount of such damage, unmanned aerial vehicles are among the best options to do the job of firefighting. Their orchestration of effective fire extinguishing is a must and essential. This work proposes a multi-agent reinforcement learning method using an actor-critic structure to extinguish the fire. While the learning is centralized, their execution is decentralized. Moreover, we define two models of fire spreading based on the discrete-time Markov process. The results show that the proposed method outperforms the state-of-the-art methods. © 2023 IEEE.

关键词： Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

Faulty Classes Prediction in Object-Oriented Programming Using Composed Dagging Technique 3

Faulty Classes Prediction in Object-Oriented Programming Usi...

引用

3rd International Conference on Electrical, computer and Communication engineering, ECCE 2023

作者： Mahfuz, Nagib Shill, Pintu Chandra North Western University Dept. of Computer Science and Engineering Khulna Bangladesh Khulna University of Engineering and Technology Dept. of Computer Science and Engineering Khulna Bangladesh

ISBN: (纸本)9798350345360

Class is one of the fundamental concepts of the object-oriented paradigm and has been scrutinized since the developers moved on from procedural programming design. In software fault prediction, the legalization of software metrics is essential. As a handful of software metrics suites exist, it is a very hard task to predict the defective classes flawlessly using a particular set of metrics suites. However, it is a rational approach to use only the object-oriented metrics that are directly relatable to the class definitions in the code that helps the developers foresee the errors in defining the classes and minimize the errors as much as possible. This paper utilized twelve object-oriented metrics selected from various metrics suites. The dagging ensemble model is merged with three well-known classification algorithms (Naive Bayes, Multilayer Perceptron, J48 Decision Tree) individually and applied to twelve java projects. The study depicts that the proposed ensemble method gives improved outcomes that are statistically significant when merged with Naive Bayes and Multilayer Perceptron. The proposed ensemble method shows improvements up to 12.5% in accuracy and 15% in F-Score. © 2023 IEEE.

关键词： Object oriented programming

来源：评论

学校读者我要写书评

暂无评论

Slang Word Detection in the Context of User Profiling in the Social Media Platformes 4

Slang Word Detection in the Context of User Profiling in the...

引用

4th International Conference on Advances in Electrical, Computing, Communication and Sustainable Technologies, ICAECT 2024

作者： Samanta, Bikram Shil, Ritika Pal, Alok Ranjan Pal, Antara College of Engg. and Management Dept. of Computer Science and Engg. Kolaghat India Atharva College of Engineering Dept. of Computer Science Engg. Mumbai India

ISBN: (纸本)9798350343670

On social media platforms, people express their views in various ways, sometimes using formal words and sometimes in informal form. The informal conversations may contain jargon words i.e. slang words which might be improper for all kinds of audience. In this work, an attempt has been made to identify slang words in a text, posted on the social media platform. The overall experiment has been carried out in 4 modules. In module-1, the suspicious words from the posted text have been identified using minimum edit distance. This experiment has been implemented by the help of a dataset of printable slang words collected from Kaggle. In module-2, the topic of discussion has been analyzed by using a knowledge based approach. This part of the experiment has been carried out by creating a newspaper archive from the online repository of the "Times of India"newspaper. But, this knowledge based strategy could not produce an appreciable accuracy at its baseline. Therefore, as a modification of this technique, in module-3, the context analysis task has been implemented combining knowledge based approach and machine learning based approach. In module-4, the module-1 and module-3 are merged to tag a suspicious word with its level of abusiveness. © 2024 IEEE.

关键词： Context Analysis Slang Word Social Media Transformer User Profiling

来源：评论

学校读者我要写书评

暂无评论

Adversarial Machine Learning Attack on Machine Learning-based Controller for Solar Inverters

Adversarial Machine Learning Attack on Machine Learning-base...

引用

2024 IEEE Energy Conversion Congress and Exposition, ECCE 2024

作者： Massa, Joaquin Kim, Kyoung-Tak Kim, Taesic Ahn, BoHyun Venayagamoorthy, Kumar Zeng, Jianwu Texas A&m University-Kingsville Dept. of Electrical Engineering and Computer Science Kingsville United States Soongsil University Dept. of Electrical Engineering Seoul Korea Republic of University of Missour-Columbia Dept. of Electrical Engineering and Computer Science Columbia United States Clemson University Dept. of Electrical and Computer Engineering Clemson United States Minnesota State University Mankato Dept. of Electrical and Computer Engineering and Technology Mankato United States

ISBN: (纸本)9798350376067

Adversarial machine learning (ML) attacks are stealthy attacks designed to mislead the ML model results. This paper explores adversarial ML attacks that generate adversarial noisy input data in an ML-based controller in a solar inverter. Three types of ML models, long short-term memory (LSTM), gated recurrent unit (GRU)), and bidirectional-LSTM (Bi-LSTM), are designed to replace proportional-integral (PI) controller-based vector control for a solar inverter and two white-box adversarial ML attacks (Basic Iterative Method (BIM) attack and Fast Sign-Gradient Method (FGSM)) are applied to the ML controllers. It is observed that the adversary ML attacks designed in stealthy way do not affect the PI controller, while significantly degrading performance of the ML-based controllers. Moreover, the BIM attack is more effective than FGSM and Bi-LSTM-based controller is relatively robust to the attacks compared to peer. © 2024 IEEE.

关键词： Adversarial machine learning

来源：评论

学校读者我要写书评

暂无评论

Predicting Mental illness (Depression) with the help of NLP Transformers 2

Predicting Mental illness (Depression) with the help of NLP ...

引用

2nd International Conference on Data science and Information System, ICDSIS 2024

作者： Kokane, Vedant Abhyankar, Ajit Shrirao, Nikhil Khadkikar, Prajakta Pune Institute Of Computer Technology Dept. Of Computer Engineering Pune India

ISBN: (纸本)9798350381665

In today's rapidly developing world, mental health is very important because it directly impacts the human thought process, productivity, relationships, etc. Being one of the most important aspects of a human's life, it is often neglected. Poor mental health can lead to anxiety, hopelessness, struggling to manage emotions, a lack of concentration, and loneliness. A mentally ill person can also have suicidal thoughts. In today's world, almost every person uses social media to share their lives and thoughts. With social media serving as a ubiquitous platform for self-expression and interaction, leveraging artificial intelligence (AI) and natural language processing (NLP) techniques becomes feasible for detecting mental health statuses based on users' social media posts. In this research paper, we proposed a methodology using NLP Transformers to study human behaviour as well as symptoms of depression. We analysed the status of mental health by examining the text of posts posted by different people on Twitter and Reddit. In our proposed methodology, we applied 4 different NLP Transformers (BERT, XLNet, RoBERTa, and DistilBERT) Transformers on two different datasets. Our proposed model concluded that DistilBert was the most accurate Transformer among the four with accuracy with 91% accuracy on twitter dataset and 84% accuracy on reddit dataset. Our proposed methodology will be effective in detecting mental illness in a person. © 2024 IEEE.

关键词： Diseases

来源：评论

学校读者我要写书评

暂无评论

Growth Rate Prediction on Fetal Brain Development using Deep Learning Techniques 5

Growth Rate Prediction on Fetal Brain Development using Deep...

引用

5th International Conference on Data Intelligence and Cognitive Informatics, ICDICI 2024

作者： Mahesh, Vellampalli Krishna Sai Kumar, Veldurthi Pavan Pandiyarajan, Pandiselvam Raveendra, Vavila Reddy, Y. Mukesh Kumar Kalasalingam Academy of Research and Education Dept. of Computer Science and Engineering Tamil Nadu India Kalasalingam Academy of Research and Education Kalasalingam University Dept. of Computer Science and Engineering Tamilnadu Krishnankovil India

ISBN: (纸本)9798350389609

The development of the fetal brain is crucial for a child's cognitive and emotional maturation, highlighting the importance of early identification of neurological issues for effective prenatal care. This study introduces an innovative deep learning framework aimed at predicting fetal brain growth rates from 2D ultrasound images. We utilize a Convolutional Neural Network (CNN) that combines longitudinal data with various imaging modalities to enhance prediction accuracy. By incorporating transfer learning alongside sophisticated techniques like attention mechanisms, our model adept.y captures both spatial and temporal dynamics within the data. Results from our experiments demonstrate that this approach outperforms conventional methods, achieving an accuracy. The findings of this research hold significant promise for prenatal healthcare by enabling timely interventions for high-risk pregnancies. Ultimately, this study contributes to a deeper understanding of fetal brain development and illustrates the potential of integrating AI technologies into standard prenatal diagnostic practices. © 2024 IEEE.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

CONVOLUTIONAL NEURAL NETWORK ALGORITHM FOR CANCELLABLE FACE TEMPLATE

UPB Scientific Bulletin, Series C: Electrical Engineering an...

引用

UPB Scientific Bulletin, Series C: Electrical engineering and computer science 2024年第1期86卷 155-166页

作者： Alwan, Hiba Basim Ku-Mahamud, Ku Ruhana Dept. of Computer Science University of Technology Iraq School of Computing Universiti Utara Malaysia Malaysia

Biometrics systems utilizing hand geometry, fingerprint, iris, face, palm print, voice, gesture, and palm print have been utilised for authentication purposes. Through these templates, the face template is suggested as the strongest. However, problems still exist in face-matching implementation which affects biometric security. Thus, in this paper, a new cancellable face image algorithm using the Convolution Neural Network (CNN) feature extraction technique and Winner-Takes-All hashing method is proposed. The algorithm is to overcome the problem of matching which will enhance its security. ORL and Yale datasets were utilized in the evaluation of the proposed algorithm. Several common algorithms were utilised in the comparison based on Equal Error Ratio (ERR) measurement. The outcomes displayed the verification performance of the unimportant differences with EER = 0.95% and 3.25% for the extended version of the face dataset for the Yale and ORL face datasets, respectively. The acceptable implementation of the proposed algorithm demonstrates that it can be used for the security purpose of biometric trait tasks. © 2024, Politechnica University of Bucharest. All rights reserved.

关键词： Convolution

来源：评论

学校读者我要写书评

暂无评论

Short-term Wind Power Ramp Forecasting Using Sequential Approach

Short-term Wind Power Ramp Forecasting Using Sequential Appr...

引用

2024 IEEE International Conference on Modeling, Simulation and Intelligent Computing, MoSICom 2024

作者： Gopalakrishnan, Leechita Barrera, Julian L Cardenas Fatima, N. Sabiyath B.S. Abdur Rahman Crescent Institute of Science and Technology Dept. of Computer Science and Engineering Chennai India University of New Brunswick Dept. of Electrical and Computer Engineering Fredericton Canada

ISBN: (纸本)9798331533311

As wind energy becomes a key player in the global transition to renewable energy, its variability presents significant challenges. Accurate wind ramp forecasting is crucial for managing such rapid fluctuations in wind power output, ensuring grid stability, and enhancing the reliability of renewable energy integration. This paper presents a short-term wind ramp forecasting model using a sequential approach: initially forecasting wind power through the integration of Variational Mode Decomposition (VMD) and the XGBoost algorithm, and subsequently detecting wind ramps using the Definition Based Sign Indicator (DSI) method. The historical wind speed and power data for Turkey was obtained from Kaggle. The dataset comprises readings from a wind turbine's SCADA system in Turkey, with measurements recorded at 10minute intervals. Variational Mode Decomposition (VMD) was applied to deconstruct and restore the historical wind power of the wind field, resulting in the formation of three training datasets. Wind power projections for the subsequent 8 hours, at 10 -minute intervals, are generated using historical wind power, wind speed, and wind direction as inputs. The DSI algorithm is employed to identify ramps. To assess the predictive capability of the proposed model, two algorithms, SVR and XGBoost, were employed for forecasting and the resulting error is minimal. The proposed model, VMD + XGBOOST is also compared without VMD to evaluate the performance. Taking mean square error (MSE), root mean square error (RMSE) mean absolute error (MAE), and symmetric mean absolute percentage error (sMAPE) as evaluation indicators, the results show that the forecasted accuracy of wind power is significantly improved after VMD processing. The performance of XGBOOST is better than SVR in the two evaluation indicators. Precision, Recall, and F1-score are calculated to evaluate the accuracy of the predicted wind ramps. © 2024 IEEE.

关键词： Wind forecasting

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：