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 ...
详细信息
ISBN:
(数字)9798350358414
ISBN:
(纸本)9798350358421
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.
This work-in-progress research-to-practice paper presents an intervention on integrating computational thinking modules into a softwareengineering course. The national consensus on the significance of computational t...
详细信息
ISBN:
(数字)9798350351507
ISBN:
(纸本)9798350363067
This work-in-progress research-to-practice paper presents an intervention on integrating computational thinking modules into a softwareengineering course. The national consensus on the significance of computational thinking has prompted the expansion of related educational initiatives over the past decade. Since the definition of computational thinking by Wing in 2006, this concept has gained significant attention within the educational community. Particularly this surge of interest has led to extensive research into its conceptual foundations and subsequent integration into educational curricula since 2013. National initiatives have since emerged to incorporate computational thinking into the educational system. Furthermore, as artificial intelligence and computing systems become increasingly integrated into daily life, there is a growing demand from industries for a workforce and graduates adept.at critical thinking and problem-solving. Aligned with this national movement, our study presents a two-year institutional initiative, aimed at integrating computational thinking into the softwareengineering program. The softwareengineering discipline extensively involves design thinking and problem-solving skills. However, we noticed that these higher-level skills are not imparted early in the program to teach students this method of thinking and approaching problems. To bridge this skill gap, we developed a set of computational thinking modules and integrated them into a gateway course in the softwareengineering program. Over two years, we implemented this intervention in an introductory-level course and evaluated its impact on students' computational thinking skills by analyzing their responses to a standard Computational Thinking Assessment survey. The results showed significant improvement in most components. These early findings underscore the effectiveness of integrating these computational thinking modules into the gateway courses, regardless of the specific co
Detection of the peripheral blood plays a vital role in the field of medical diagnostics and control of major diseases. In this regard, detecting leukocytes or White Blood Cells (WBCs) is important. Leukocytes have fi...
详细信息
ISBN:
(数字)9798350379587
ISBN:
(纸本)9798350379594
Detection of the peripheral blood plays a vital role in the field of medical diagnostics and control of major diseases. In this regard, detecting leukocytes or White Blood Cells (WBCs) is important. Leukocytes have five classes: Basophils, Eosinophils, Lymphocytes, Monocytes, and Neutrophils. The dataset used for this study is Raabin-WBC, containing leukocyte images. Deep learning techniques are applied to the dataset for the detection of mentioned leukocyte classes. In this case, to remove class imbalance, oversampling/replication of data is done. Transfer learning is applied in which a pre-trained ResNet18 model gets adapted to new data. Deep features are extracted from the pool5 layer of the network and are fed to the Ensemble Bagged Trees classifier. A remarkable classification accuracy of 99.5% is achieved through 5-fold cross-validation.
Opioid abuse and dependence have emerged as a pressing global concern, posing significant challenges to public health and society. Early identification and prediction of opioid dependency represent crucial steps in mi...
Opioid abuse and dependence have emerged as a pressing global concern, posing significant challenges to public health and society. Early identification and prediction of opioid dependency represent crucial steps in mitigating its abuse impact on individuals and communities. The application of machine learning techniques to analyze medical data has opened new avenues for achieving this goal. While this field and the prediction is still in its infancy, our research explores the potential of several machine learning algorithms including LightGBM for this risk prediction. To tackle the inherent class imbalance in the MIMICIII dataset, we implemented the Synthetic Minority Oversampling Technique (SMOTE). We developed predictive models using four distinct algorithms: decision trees, random forests, support vector machines, and LightGBM. These models were meticulously evaluated to assess their performance. Ultimately, our findings revealed that the LightGBM model outperformed the other algorithms, demonstrating superior accuracy and achieving a higher Area Under the Curve value. This outcome underscores the potential of LightGBM as a valuable algorithm in the early prediction of the risk of opioid dependence, thereby offering substantial benefits to both patients and society at large.
The paper proposes a cryptographic protocol two-factor authentication with the zero-knowledge over the extended field GF(2 m ) on elliptic curves using biometric data and private key of the user. The implementation of...
The paper proposes a cryptographic protocol two-factor authentication with the zero-knowledge over the extended field GF(2 m ) on elliptic curves using biometric data and private key of the user. The implementation of a cryptographic protocol with zero-knowledge proof based on elliptic curves allows significantly reducing the size of protocol parameters and increasing the cryptographic strength (computational complexity of the breaking). The cryptographic protocol was modeled in the High-Level Protocol Specification Language, the model validation and protocol verification was performed using the Security Protocol Animator tool for Automated Validation of Internet Security Protocols and Applications. The software verification of the cryptographic protocol was performed using the software modules On the Fly Model Checker and Constraint Logic based Attack Searcher.
The paper demonstrates the process of developing mathematical models for identifying breakdowns of electric motors using machine learning methods. The authors have developed three mathematical models for identifying b...
详细信息
This research project proposes the use of interactive videos to enhance teachers' training in educating autistic children. Effective teaching methods for students with autism require teachers to comprehend the con...
This research project proposes the use of interactive videos to enhance teachers' training in educating autistic children. Effective teaching methods for students with autism require teachers to comprehend the condition and employ tailored instructional strategies, including adapting assignments, aiding those with language difficulties, and utilizing visual aids for better organization and focus. Traditional teacher training methods can be both expensive and time-consuming. In contrast, interactive videos provide a proactive and flexible way to access training content, empowering teachers to engage with the material dynamically and take control of their learning experiences. Future work will explore the integration of AI-driven ChatGPT to offer personalized support and create a dynamic training program, with the goal of improving inclusivity and educational quality in autism settings while benefiting teachers, students, and the education system at large.
In a world with an overgrowing elderly population, there exists a critical need for a greater number of skilled individuals in the nursing industry. AI-based systems can be useful, compared to traditional ones with re...
详细信息
ISBN:
(数字)9798350375503
ISBN:
(纸本)9798350375510
In a world with an overgrowing elderly population, there exists a critical need for a greater number of skilled individuals in the nursing industry. AI-based systems can be useful, compared to traditional ones with require in-person assistance, to accurately identify nursing activities and assess the nursing trainees to help them become proficient. This paper addresses classifying activities in one such procedure, endotracheal suctioning, using skeletal keypoint data of the subject performing the procedure. A multi-step structured prompt engineering method was established and utilized on several LLMs to select or calculate key features from the data. Then the features are passed onto several tuned machine learning models to obtain results. A tuned XGBoost prevailed across all models, achieving 90% accuracy on the validation set.
This paper presents a quantitative analysis of trends in Russian literature and Korean-Russian translated literature within Korean academia from 1996 to 2024. By analyzing frequency themes and contexts, the study expl...
详细信息
ISBN:
(数字)9798350367874
ISBN:
(纸本)9798350367881
This paper presents a quantitative analysis of trends in Russian literature and Korean-Russian translated literature within Korean academia from 1996 to 2024. By analyzing frequency themes and contexts, the study explores evolving research trends, revealing how sociocultural, political, and artistic exchanges between Korea and Russia have influenced each other. Future research using text mining promises to uncover more specific patterns, offering deeper insights into literary and cultural exchanges. This study contributes to understanding the role of literature in expressing and understanding sociopolitical and cultural values in both countries.
DNA sequence classification is a major challenge in biological processing of data. The classification of DNA sequences is an important study field in bioinformatics since it allows researchers to perform genomic analy...
详细信息
ISBN:
(数字)9798350379716
ISBN:
(纸本)9798350379723
DNA sequence classification is a major challenge in biological processing of data. The classification of DNA sequences is an important study field in bioinformatics since it allows researchers to perform genomic analysis. Through this research, we tried to analyze the performance of various machine learning (ML) algorithms in identifying DNA genes in humans and chimpanzees to find the most effective algorithm for classification of DNA sequences. The results of this research outperformed earlier studies, which had an accuracy around 98%. The dataset was preprocessed using k-mer counting, which assisted in finding key characteristics for the model's training. Naïve Bayes stood out, with an outstanding accuracy of 98.4% for human DNA and 91.4% for chimpanzee DNA. After improving the parameters, Naïve Bayes continued to improve, with human DNA accuracy increasing to 99% and chimpanzee DNA accuracy to 92% in all categories followed by Logistic Regression at 93.9% for human DNA sequences and Ridge Classifier at 90.2% for chimpanzee DNA sequences. K-Nearest Neighbors, Ridge Classifier, Logistic Regression, Decision Tree, Random Forest, Gradient Boosting, Extra Trees, XGBoost, and Naïve Bayes were among the algorithms tested, with accuracy, F1 score, precision, and recall used in assessing their performance. These findings highlight the necessity of selecting the appropriate models and tweaking their parameters to improve performance.
暂无评论