检索结果-内蒙古大学图书馆

Parameter-efficient fine-tuning of pre-trained code models for just-in-time defect prediction

Neural Computing and Applications 2024年第27期36卷 16911-16940页

作者： Abu Talib, Manar Bou Nassif, Ali Azzeh, Mohammad Alesh, Yaser Afadar, Yaman Department of Computer Science College of Computing and Informatics University of Sharjah Sharjah United Arab Emirates Department of Computer Engineering College of Computing and Informatics Sharjah United Arab Emirates King Hussein School for Computing Sciences Princess Sumaya University for Technology Amman Jordan

Software engineering workflows use version control systems to track changes and handle merge cases from multiple contributors. This has introduced challenges to testing because it is impractical to test whole codebases to ensure each change is defect-free, and it is not enough to test changed files alone. Just-in-time software defect prediction (JIT-SDP) systems have been proposed to solve this by predicting the likelihood that a code change is defective. Numerous techniques have been studied to build such JIT software defect prediction models, but the power of pre-trained code transformer language models in this task has been underexplored. These models have achieved human-level performance in code understanding and software engineering tasks. Inspired by that, we modeled the problem of change defect prediction as a text classification task utilizing these pre-trained models. We have investigated this idea on a recently published dataset, ApacheJIT, consisting of 44k commits. We concatenated the changed lines in each commit as one string and augmented it with the commit message and static code metrics. Parameter-efficient fine-tuning was performed for 4 chosen pre-trained models, JavaBERT, CodeBERT, CodeT5, and CodeReviewer, with either partially frozen layers or low-rank adaptation (LoRA). Additionally, experiments with the Local, Sparse, and Global (LSG) attention variants were conducted to handle long commits efficiently, which reduces memory consumption. As far as the authors are aware, this is the first investigation into the abilities of pre-trained code models to detect defective changes in the ApacheJIT dataset. Our results show that proper fine-tuning improves the defect prediction performance of the chosen models in the F1 scores. CodeBERT and CodeReviewer achieved a 10% and 12% increase in the F1 score over the best baseline models, JITGNN and JITLine, when commit messages and code metrics are included. Our approach sheds more light on the abilities of l

关键词： Forecasting

来源：评论

学校读者我要写书评

暂无评论

Recognition of Dog Motion States: Ensemble vs Deep Learning Models 15

Recognition of Dog Motion States: Ensemble vs Deep Learning ...

引用

15th International Conference on Information, Intelligence, Systems and Applications, IISA 2024

作者： Davoulos, George Lalakou, Iro Hatzilygeroudis, Ioannis University of Patras Computer Engineering and Informatics Department Patras Greece University of Peloponnese Department of Informatics and Telecommunications Tripolis Greece

ISBN: (纸本)9798350368833

Dog activities recognition, especially dog motion status recognition, is an active research area. Although several machine learning and deep learning approaches have been used for dog motion states recognition, use of ensemble learning methods is rather missing as well as a comparison with deep learning ones. This paper focuses on the use of ensemble classifiers and deep learning neural networks in recognizing dog motion states, and their comparison. Our results show a slight superiority in accuracy (94.7% vs 92.8%) of deep learning, which is quite comparable to the state-of-the-art, but at the cost of larger complexity and training time, which makes ensemble techniques still attractive. © 2024 IEEE.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

Frame-by-Frame Analysis for Assessing Chickens Flock Movement 25

Frame-by-Frame Analysis for Assessing Chickens Flock Movemen...

引用

25th International Electronics Symposium, IES 2023

作者： Sesulihatien, Wahjoe Tjatur Barakbah, Ali Ridho Prasetyo, Piko Permata Ilham Dept. of Informatics and Computer Engineering Surabaya Indonesia

ISBN: (纸本)9798350314731

Assessing the behavior and welfare of chickens is crucial for successful chicken rearing and product quality. Recent research has focused on using advanced cameras and computer vision technologies to analyze chicken behavior. This camera-based approach captures high-resolution video footage, which is then processed using computer vision algorithms to extract insights on movement patterns, feeding behavior, and social interactions. While this approach has advanced our understanding, there is a need for studies employing a frame-by-frame analysis to obtain fine-grained details of chicken behavior. This method allows for precise measurements and real-Time analysis, facilitating informed decision-making in chicken farming. To expedite the assessment process, a bandpass filter 470 nm lens is incorporated into the camera setup, optimizing image segmentation and reducing processing time. The proposed system combines this specialized camera with frame-by-frame calculations to track and measure the velocity, acceleration, and position of chickens within a flock. The integration of these technologies holds the potential for improving our understanding of chicken behavior, enhancing welfare, and driving data-driven decision-making in chicken farming practices. © 2023 IEEE.

关键词： computer vision

来源：评论

学校读者我要写书评

暂无评论

Identification of Software Bugs by Analyzing Natural Language-Based Requirements Using Optimized Deep Learning Features

引用

computers, Materials & Continua 2024年第3期78卷 4379-4397页

作者： Qazi Mazhar ul Haq Fahim Arif Khursheed Aurangzeb Noor ul Ain Javed Ali Khan Saddaf Rubab Muhammad Shahid Anwar Department of Computer Science and Engineering and International Bachelor’s Program in Informatics Yuan Ze UniversityZhongliTaoyuan320315Taiwan Department of Computer Science National University of PakistanRawalpindiPunjab46000Pakistan Department of Computer Software Engineering MCSNational University of Sciences and TechnologyIslamabad44000Pakistan Department of Computer Engineering College of Computer and Information SciencesKing Saud UniversityP.O.Box 51178Riyadh11543Saudi Arabia Department of Computer Science School of PhysicsEngineering and Computer ScienceUniversity of HertfordshireHatfieldAL109ABUK Department of Computer Engineering College of Computing and InformaticsUniversity of SharjahSharjah27272United Arab Emirates Department of AI and Software Gachon UniversitySeongnam-SI13120South Korea

Software project outcomes heavily depend on natural language requirements,often causing diverse interpretations and issues like ambiguities and incomplete or faulty *** are exploring machine learning to predict software bugs,but a more precise and general approach is *** bug prediction is crucial for software evolution and user training,prompting an investigation into deep and ensemble learning ***,these studies are not generalized and efficient when extended to other ***,this paper proposed a hybrid approach combining multiple techniques to explore their effectiveness on bug identification *** methods involved feature selection,which is used to reduce the dimensionality and redundancy of features and select only the relevant ones;transfer learning is used to train and test the model on different datasets to analyze how much of the learning is passed to other datasets,and ensemble method is utilized to explore the increase in performance upon combining multiple classifiers in a *** National Aeronautics and Space Administration(NASA)and four Promise datasets are used in the study,showing an increase in the model’s performance by providing better Area Under the Receiver Operating Characteristic Curve(AUC-ROC)values when different classifiers were *** reveals that using an amalgam of techniques such as those used in this study,feature selection,transfer learning,and ensemble methods prove helpful in optimizing the software bug prediction models and providing high-performing,useful end mode.

关键词： Natural language processing software bug prediction transfer learning ensemble learning feature selection

来源：评论

学校读者我要写书评

暂无评论

Screen Reading Regions in Social Media Comments: An Eye-Tracking Analysis of Visual Attention on Smartphones 27

Screen Reading Regions in Social Media Comments: An Eye-Trac...

引用

27th Pan-Hellenic Conference on Progress in Computing and informatics with International Participation, PCI 2023

作者： Dritsa, Stavroula Mallas, Andreas Xenos, Michalis University of Patras Department of Computer Engineering and Informatics Greece

ISBN: (纸本)9798400716263

Our study examines how formatted text influences reading behavior concerning the smartphone screen regions users utilize on social media comments. Aiming to research visual attention in scrollable content, we investigate the differences in the distribution of visual attention in screen regions between social media comment sections and the users' preferred reading regions across all cases examined. An experiment was conducted with participants (n = 47) engaging in reading activities on the comment section of four social media -Twitter, YouTube (in two versions), Facebook, and Instagram- chosen due to their popularity and the users' familiarity with them. Results showed that users' visual attention is distributed differently between social media. Different reading styles were observed, with users not utilizing the entirety of the screen but instead focusing on specific screen regions that varied between users. © 2023 Owner/Author.

关键词： Smartphones

来源：评论

学校读者我要写书评

暂无评论

Sentiment Analysis on the Use of Electric Vehicles With Long Short-Term Memory

Sentiment Analysis on the Use of Electric Vehicles With Long...

引用

2024 International Conference on Control and Automation, Electronics, Robotics, Internet of Things, and Artificial Intelligence, CERIA 2024

作者： Nalawati, Rizki Elisa Liliana, Dewi Yanti Iswara, Ratna Widya Rozana, Izzati Fauziyah, Putri Dzakiyah Politeknik Negeri Jakarta Dept. of Informatics and Computer Engineering Depok Indonesia

ISBN: (纸本)9798331511166

Electric vehicles, or EVs, have drawn a lot of attention lately as an eco-friendly way to cut carbon emissions and lessen reliance on fossil fuels. This work uses a neural network model called Long Short-Term Memory (LSTM), which is effective at processing sequential input, to analyze user attitudes toward electric vehicles. Opinion data were gathered from the social media network Twitter between March 1st and March 31st, 2023, using the keyword "electric vehicles". Based on the obtained opinion data, the LSTM approach is utilized to categorize sentiments as positive or negative. Implementation of the random oversampling technique is important to be applied in increasing the accuracy of the n recall model for the specified class. The study's conclusions show that the LSTM model performs admirably when it comes to sentiment analysis using the test dataset. The model evaluation yields a precision of 99.65% and an accuracy rating of 97.79%. The modeling process is carried out with variance splitting data where the highest accuracy is obtained from a comparison of training testing data of 90%: 10%. © 2024 IEEE.

关键词： Long short-term memory

来源：评论

学校读者我要写书评

暂无评论

CNN Architectures for Vehicle Detection in V2X Environments: A Comparative Study

CNN Architectures for Vehicle Detection in V2X Environments:...

引用

2024 IEEE Conference on Standards for Communications and Networking, CSCN 2024

作者： Efthymiou, Vasileios Basagiannis, Stylianos Petridou, Sophia University of Macedonia Department of Applied Informatics Greece International Hellenic University Department of Computer Informatics & Telecommunications Engineering Greece

ISBN: (纸本)9798331507428

Leveraging convolutional neural networks (CNNs) to Vehicle-to-Everything (V2X) communication systems offers a promising approach to vehicle detection, which is essential for improving road safety and traffic efficiency. At the same time, parameterization of CNNs' architectures in order to achieve high accuracy in moving objects' detection becomes an engineering challenge, especially when data pre- and post-processing solutions need to respect timing requirements. Our study evaluates the performance of five CNN architectures, namely YOLO in v5, v8, and v9, Faster R-CNN, and VGG16, for car detection. We utilize an open dataset including images derived from road-traffic environments and assess the models' performance using benchmark evaluation metrics, i.e., precision, recall, mean Average Precision (mAP50, mAP50 - 95), and F1-score. Furthermore, we analyze their computational requirements, in terms of execution time and resource demands, which are critical factors when considering real-time applications in V2X systems, such as autonomous driving and traffic management. Our evaluation results indicate that, while the YOLO architecture exhibits the best balance of speed and accuracy, Faster R-CNN has higher detection performance at the cost of greater computational demand. This paper provides insights into each model's strengths and limitations offering guidance for selecting optimal architectures tailored to V2X-related applications. © 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

Current Trends, Challenges and Techniques in XAI Field;A Tertiary Study of XAI Research 47

Current Trends, Challenges and Techniques in XAI Field;A Ter...

引用

47th ICT and Electronics Convention, MIPRO 2024

作者： Brdnik, Saša Šumak, Boštjan University Of Maribor Institute Of Informatics Faculty Of Electrical Engineering And Computer Science Slovenia

ISBN: (纸本)9798350382495

This tertiary systematic literature review examines 29 systematic literature reviews and surveys in Explainable Artificial Intelligence (XAI) to uncover trends, limitations, and future directions. The study explores current explanation techniques, providing insights for researchers, practitioners, and policymakers interested in enhancing AI transparency. Notably, the increasing number of systematic literature reviews (SLRs) in XAI publications indicates a growing interest in the field. The review offers an annotated catalogue for human-computer interaction-focused XAI stakeholders, emphasising practitioner guidelines. Automated searches across ACM, IEEE, and Science Direct databases identified SLRs published between 2019 and May 2023, covering diverse application domains and topics. While adhering to methodological guidelines, the SLRs often lack primary study quality assessments. The review highlights ongoing challenges and future opportunities related to XAI evaluation standardisation, its impact on users, and interdisciplinary research on ethics and GDPR aspects. The 29 SLRs, analysing over two thousand papers, include five directly relevant to practitioners. Additionally, references from the SLRs were analysed to compile a list of frequently cited papers, serving as recommended foundational readings for newcomers to the field. © 2024 IEEE.

关键词： Artificial intelligence

来源：评论

学校读者我要写书评

暂无评论

Evaluation of a Serious Game for Young Software engineering Graduates: Interviewing Skills and Soft Skills 14

Evaluation of a Serious Game for Young Software Engineering ...

引用

14th IEEE Global engineering Education Conference, EDUCON 2023

作者： Evangelou, Semira Maria Kapsoulakis, Vangelis Xenos, Michalis University of Patras Computer Engineering and Informatics Patras Greece

ISBN: (纸本)9798350399431

Serious Games have proved to have instructional potential due to their ability to present realistic simulations of real-life situations. In this study, a serious game called 'Galaxy Solutions' was developed. This game simulates all the stages of the recruitment process. It starts when the player is searching for vacancies and submitting applications, continues to the interview process, and concludes at the last stage, in which a company named 'Jupiter Solutions' hires the player. Consequently, new challenges unfold in the working environment. The main objective of this game is to help graduates acquire basic interview skills and be aware of the whole process by making the learning experience more enjoyable and engaging. Throughout the game, the player must make decisions through multiple-choice actions and dialogue, and all these decisions affect the character's events in the future. The game was evaluated by 105 students, and it was found usable by all the students, regardless of their success level in the game, and had a statistically significant learning effect, as the pre-posttest analysis revealed. © 2023 IEEE.

关键词： Serious games

来源：评论

学校读者我要写书评

暂无评论

Comparison of Q&A Classification Using BiLSTM Based on Closed Domain Knowledge 12

Comparison of Q&A Classification Using BiLSTM Based on Close...

引用

12th Electrical Power, Electronics, Communications, Controls and informatics Seminar, EECCIS 2024

作者： Gunawan, Alifi Lazuardi Bachtiar, Fitra Abdurrachman Setiawan, Budi Darma University Malang Faculty of Computer Science Brawijaya Informatics Engineering Department Indonesia

ISBN: (纸本)9798350379914

The Faculty of computer Science, Universitas Brawijaya (Filkom UB) is committed to providing quality services for the users especially internal and external stakeholders, one of which is through the HaloFilkom service. HaloFilkom services have limitations in terms of time. HaloFilkom services are not available 24 hours due to limited working hours. Questions asked by users are not answered directly. This weakness in the HaloFilkom system can be overcome by using a chatbot system. Chatbot is an interactive system that works with natural human language and can work 24 hours. Thus, the current study explores the basic chatbot model by classifying the Q&A in the closed domain knowledge. The dataset in this research is in the form of pairs of questions and answers regarding various topics at the Filkom UB. The knowledge is preprocessed using text preprocessing which includes case folding, tokenization, padding, and tensorization. One of the chatbot models is a generative model. Creating a generative chatbot model can be done using the Seq2Seq model mechanism which consists of an encoder and decoder. The model created consists of four different architectures, namely a model with an LSTM encoder without attention and with attention and a BiLSTM model encoder without attention and with attention. Hyperparameter tuning was conducted to obtain the best hyperparameter combination. The experiment results show the best hyperparameter combination obtained is hidden size 448, drop out rate 0.5, learning rate 0.001, batch size 64, and teacher force 0. The model with the best loss is obtained with a BiLSTM encoder architecture without an attention mechanism with a train loss of 0.120. The model with the highest BLEU Score was obtained by a model with a BiLSTM encoder architecture without an attention mechanism with a BLEU Score of 0.8587 on the training data. Testing using prompt testing obtained an average BLEU Score of 0.3745 on the BiLSTM encoder without an attention mechanism mo

关键词： Teaching

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：