Article
Neurosciences
Thomas Jochmann, Marc S. Seibel, Elisabeth Jochmann, Sheraz Khan, Matti S. Haemaelaeinen, Jens Haueisen
Summary: This study investigates a convolutional neural network that detects sex from clinical EEG and finds that electrocardiac artifacts leak into the classifier. However, even after removing these artifacts, the sex can still be determined from the EEG, with topographies being critical but waveforms and frequencies not important for sex detection.
HUMAN BRAIN MAPPING
(2023)
Article
Computer Science, Artificial Intelligence
Jue Wang, Anoop Cherian
Summary: This research introduces a discriminative pooling approach that utilizes support vector machine multi-instance learning to learn a separating hyperplane based on features from positive and negative bags, achieving effective description of actions in videos.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2021)
Article
Chemistry, Analytical
Huaqing Wang, Zhitao Xu, Xingwei Tong, Liuyang Song
Summary: The application of transfer learning in fault diagnosis has been developed in recent years to solve the problem of fault recognition under different working conditions. The open set recognition ability of the transfer learning method is an urgent research direction due to the changeable status of equipment and the collection of signals with new fault classes.
Article
Oncology
Roman C. Maron, Achim Hekler, Sarah Haggenmueller, Christof von Kalle, Jochen S. Utikal, Verena Mueller, Maria Gaiser, Friedegund Meier, Sarah Hobelsberger, Frank F. Gellrich, Mildred Sergon, Axel Hauschild, Lars E. French, Lucie Heinzerling, Justin G. Schlager, Kamran Ghoreschi, Max Schlaak, Franz J. Hilke, Gabriela Poch, Soren Korsing, Carola Berking, Markus Heppt, Michael Erdmann, Sebastian Haferkamp, Dirk Schadendorf, Wiebke Sondermann, Matthias Goebeler, Bastian Schilling, Jakob N. Kather, Stefan Froehling, Daniel B. Lipka, Eva Krieghoff-Henning, Titus J. Brinker
Summary: This study investigates the performance of model soups for skin cancer classification and finds that they improve generalization, robustness, and calibration.
EUROPEAN JOURNAL OF CANCER
(2022)
Review
Biochemistry & Molecular Biology
Jaroslaw Polanski
Summary: The availability of computers has opened new possibilities in drug design, and neural networks have played an important role. However, with the recent success of deep learning, there has been a resurgence in the use of neural networks in deep chemistry. Self-organizing maps have been found to be highly efficient in molecular representation. While deep learning has shown efficiency in other areas, its application in deep chemistry still faces challenges.
INTERNATIONAL JOURNAL OF MOLECULAR SCIENCES
(2022)
Article
Neurosciences
Huaiqiang Sun, Guoting Luo, Su Lui, Xiaoqi Huang, John Sweeney, Qiyong Gong
Summary: This study proposes a specially designed autoencoder to investigate structural brain changes in schizophrenia. The classifier trained with autoencoded features outperforms the classifier trained with conventional morphological features in identifying schizophrenia patients from healthy controls.
HUMAN BRAIN MAPPING
(2023)
Article
Biochemistry & Molecular Biology
Jonathan Kim, Stefan Bekiranov
Summary: Quantum metric learning is a method that can classify data into categories by learning through quantum embedding, achieving classification in high-dimensional feature data. The research found that by reducing data dimensionality and limiting the number of model parameters, quantum metric learning can accurately classify test data.
Article
Chemistry, Analytical
Wenlang Xie, Zhixiong Li, Yang Xu, Paolo Gardoni, Weihua Li
Summary: Bearing fault diagnosis is crucial in aerospace, marine, and heavy industries for improving machine life, reducing economic losses, and preventing safety problems. Traditional methods for fault feature extraction face challenges, while deep neural networks can automatically extract intrinsic features. This study built four hybrid models based on CNN and compared their fault detection accuracy and efficiency. The results showed that random forest and support vector machine can maximize the CNN feature extraction ability.
Article
Oncology
Nazik Alturki, Muhammad Umer, Abid Ishaq, Nihal Abuzinadah, Khaled Alnowaiser, Abdullah Mohamed, Oumaima Saidani, Imran Ashraf
Summary: This study proposes a hybrid model for brain tumor detection using features extracted from a convolutional neural network. Experimental results show that the model achieves a 99.9% accuracy in detecting brain tumors, outperforming manually extracted features. Early detection of brain tumors is crucial for effective treatment, as they are among the top ten leading fatal diseases.
Review
Computer Science, Artificial Intelligence
Pengzhi Li, Yan Pei, Jianqiang Li
Summary: Autoencoder is an unsupervised learning model that automatically learns data features and acts as a dimensionality reduction method. This paper explains the principle and development process of a conventional autoencoder, proposes a taxonomy of autoencoders based on their structures and principles, analyzes and discusses various autoencoder models, introduces their applications in different fields, and summarizes the shortcomings of the current autoencoder algorithm while addressing future development directions.
APPLIED SOFT COMPUTING
(2023)
Article
Radiology, Nuclear Medicine & Medical Imaging
Yee Liang Thian, Dian Wen Ng, James Thomas Patrick Decourcy Hallinan, Pooja Jagmohan, Soon Yiew Sia, Jalila Sayed Adnan Mohamed, Swee Tian Quek, Mengling Feng
Summary: This study investigates the effect of training dataset size on the performance of deep learning classifiers in the radiology domain, using chest radiograph pneumothorax detection as a proxy visual task. The results indicate that there is a point of diminishing performance returns for increasing training data volumes, which has important implications for the high costs of obtaining and labelling radiology data.
JOURNAL OF DIGITAL IMAGING
(2022)
Article
Optics
Yan Teng, Chun Li, Shaochen Li, Yuhua Xiao, Ling Jiang
Summary: In this study, a method for designing terahertz random meta-surfaces based on deep Convolutional Neural Networks and genetic algorithms is proposed. The forward prediction model accurately predicts the reflection amplitude and phase response, with a calculation speed increased by 40,000 times compared to the Full-wave solver. By combining with genetic algorithms, the efficiency of the design is greatly improved, providing an efficient method for global optimization in complex designs.
OPTICS AND LASER TECHNOLOGY
(2023)
Article
Computer Science, Artificial Intelligence
Giuseppe C. Calafiore, Giulia Fracastoro
Summary: This article introduces two novel sparse versions of the classical nearest-centroid classifier, which perform simultaneous feature selection and classification at a linear computational cost. The proposed classifiers select the most relevant features and have low complexity for testing and classifying new samples.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Computer Science, Interdisciplinary Applications
Alexandre Bailly, Corentin Blanc, Elie Francis, Thierry Guillotin, Fadi Jamal, Bechara Wakim, Pascal Roy
Summary: This study compares the impact of training dataset size and interactions on the performance of machine learning and deep learning models. The results show that machine learning models are less influenced by dataset size but require interaction terms to achieve good performance, while deep learning models can achieve good performance even without interaction terms. Overall, well-specified machine learning models outperform deep learning models.
COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE
(2022)
Article
Computer Science, Artificial Intelligence
Yanchen Liu, King Wai Chiu Lai
Summary: This paper proposes an evaluation index and algorithm for addressing the common problem of class imbalance in classification domains. The Model Performance Index (MPI) is introduced as a new evaluation metric considering the impact of class imbalance. The Ideal Model Performance Algorithm is developed to estimate ideal classifier performance with a fair distribution. Experimental results show that MPI is more sensitive than traditional metrics in detecting changes in classifier performance, and the algorithm achieves small differences between predicted and observed values. It demonstrates practical value in estimating and representing classifier performances.
PATTERN RECOGNITION
(2023)
Article
Chemistry, Multidisciplinary
Kaixiang Du, Mingyuan Wang, Zhiping Liang, Quanjiang Lv, Haigang Hou, Shuangying Lei, Shahid Hussain, Guiwu Liu, Junlin Liu, Guanjun Qiao
Summary: This paper presents a strategy for synthesizing surface-passivated monocrystalline PNRs on a silicon substrate using space-confined chemical vapor transport. The growth mechanism of the PNRs is revealed, showing that P-4 molecules can break, restructure, and nucleate on the surface of the Au3SnP7 catalyst, preferentially growing along the zigzag direction to form PNRs. The growth of PNRs with structural integrity can be regulated by controlling the phosphorus molecule concentration and confined space.
CHEMICAL COMMUNICATIONS
(2023)
Article
Environmental Sciences
Jesse Nii Okai Amu-Darko, Shahid Hussain, Xiangzhao Zhang, Asma A. Alothman, Mohamed Ouladsmane, M. Tariq Nazir, Guanjun Qiao, Guiwu Liu
Summary: Through thermal treatment, a mixed heterostructures In2O3/ZnO hollow nanocages were prepared, which showed exceptional sensitivity and selectivity to hydrogen sulfide gas, greatly improving the gas sensing performance.
Article
Computer Science, Software Engineering
Fuyang Li, Wanpeng Lu, Jacky Wai Keung, Xiao Yu, Lina Gong, Juan Li
Summary: Effort-Aware Defect Prediction (EADP) methods rank software modules based on defect density and prioritize inspection of high-density modules. However, the impact of feature selection methods on EADP performance is unknown. This study examined 24 feature selection methods with 10 classifiers in a state-of-the-art EADP model on 41 PROMISE defect datasets. The results show that wrapper-based methods with forward search perform best, particularly XGBF with XGBoost as the embedded classifier in CBS+.
Article
Computer Science, Information Systems
Xiao Yu, Heng Dai, Li Li, Xiaodong Gu, Jacky Wai Keung, Kwabena Ebo Bennin, Fuyang Li, Jin Liu
Summary: This study aims to find a stable ranking of learning to rank algorithms to determine the best ones for software projects. By conducting experiments on multiple datasets and evaluating using multiple metrics based on modules and lines of code, the best algorithms were determined.
INFORMATION AND SOFTWARE TECHNOLOGY
(2023)
Article
Computer Science, Software Engineering
Fengji Zhang, Jin Liu, Yao Wan, Xiao Yu, Xiao Liu, Jacky Keung
Summary: Stack Overflow is a popular programming community where developers can seek help. To help developers describe their problems more effectively and get the answers they anticipate, we propose M3NSCT5, a novel approach for automatically generating multiple post titles from code snippets.
JOURNAL OF SYSTEMS AND SOFTWARE
(2023)
Article
Physics, Condensed Matter
Muhammad Umer Iqbal, Sumayya, Sajid Butt, Muhammad Umer Farooq, Shahid Hussain, Syed Irfan, Nazakat Ali, Muhammad Abdul Basit, Muhammad Aftab Akram, Muhammad Yasir, Ather Hassan
Summary: During the last decade, misfit layered calcium cobaltite (Ca3Co4O9) has been explored as a potential thermoelectric material. In this study, the thermoelectric properties of Ca3Co4O9 were improved by synthesizing composites with different wt.% of Cu. The composites were consolidated into bulk ceramics and their structure was analyzed, revealing a multiphase system. The sample containing 5% Cu demonstrated the highest power factor at 973K, which resulted from the simultaneous improvement in electrical conductivity and Seebeck coefficient. The conduction mechanisms were further discussed.
PHYSICA B-CONDENSED MATTER
(2023)
Article
Chemistry, Analytical
Khalid Saeed, Wajeeha Khalil, Ahmad Sami Al-Shamayleh, Iftikhar Ahmad, Adnan Akhunzada, Salman Z. ALharethi, Abdullah Gani
Summary: The exponentially growing concern of cyber-attacks on extremely dense underwater sensor networks (UWSNs) and the evolution of UWSNs digital threat landscape has brought novel research challenges and issues. This research implements an active attack in the Adaptive Mobility of Courier Nodes in Threshold-optimized Depth-based Routing (AMCTD) protocol to evaluate its performance. The preliminary research findings show that active attack drastically lowers the AMCTD protocol's performance.
Article
Green & Sustainable Science & Technology
Khalid Saeed, Wajeeha Khalil, Ahmad Sami Al-Shamayleh, Sheeraz Ahmed, Adnan Akhunzada, Salman Z. Alharthi, Abdullah Gani
Summary: This research analyzes the security-based schemes in underwater wireless sensor networks (UWSNs) and categorizes them into five sub-categories. It discusses the major contributions, techniques used, possible future research issues, and implementation tools for each security-based scheme. The identified open research issues and future trends can be further explored by the research community.
Editorial Material
Computer Science, Information Systems
Wai Keung Jacky, Leonardo Mariani, Jianwen Xiang, Xiao Yu
INFORMATION AND SOFTWARE TECHNOLOGY
(2023)
Article
Chemistry, Multidisciplinary
Shahid Hussain, Samuel B. Adeloju
Summary: This study describes the fabrication of a highly selective and ultrasensitive sulfite nanobiosensor using a layered architectural fabrication method. The nanobiosensor exhibits fast response time, wide linear calibration range, and excellent sensitivity, and can be successfully applied to sulfite determination in real samples.
Article
Engineering, Environmental
Jesse Nii Okai Amu-Darko, Shahid Hussain, Qiang Gong, Xiangzhao Zhang, Ziwei Xu, Mingsong Wang, Guiwu Liu, Guanjun Qiao
Summary: In this study, In2O3/PANI composites were synthesized using the hydrothermal technique to explore their sensing capabilities for NO2 gas. The gas sensing properties, such as response/recovery time, optimal temperature, and recyclability, were investigated by analyzing the morphology, purity, and crystal structures of the samples. The In2O3/PANI-1 sensors exhibited high responses, quick response and recovery times, and a low detection limit to NO2 gas at a working temperature of 250 degrees C. The fundamental detecting mechanism for NO2 gas and the electrical characteristics of the In2O3/PANI composites were also discussed.
JOURNAL OF ENVIRONMENTAL CHEMICAL ENGINEERING
(2023)
Article
Engineering, Multidisciplinary
Shahid Hussain, Kazi Sabiruddin
Summary: Different phases formed during the synthesis of hydroxyapatite (HA) powder by hydrothermal reaction via Indian clam seashell due to incomplete reaction. Rietveld refinement method was used to calculate the weight fraction of different phases. The highest amount of HA phase was found in powders synthesized at 700, 800, and 900 degrees C for 2 h duration, while highly pure HA powders can be prepared at 1000 degrees C and 1100 degrees C for 3 h time. Raman spectroscopy was used to study the vibration modes of PO43- tetrahedral.
SADHANA-ACADEMY PROCEEDINGS IN ENGINEERING SCIENCES
(2023)
Article
Computer Science, Information Systems
Zhen Yang, Jacky Wai Keung, Zeyu Sun, Yunfei Zhao, Ge Li, Zhi Jin, Shuo Liu, Yishu Li
Summary: This paper presents MetaCoder, a meta-learning code generation approach that efficiently extracts general-purpose knowledge from large-scale source languages and rapidly adapts to domain-specific scenarios.
INFORMATION AND SOFTWARE TECHNOLOGY
(2024)
Article
Computer Science, Information Systems
Zulfiqar Ali, Asif Muhammad, Ahmad Sami Al-Shamayleh, Kashif Naseer Qureshi, Wagdi Alrawagfeh, Adnan Akhunzada
Summary: This study introduces a deep learning-based movie recommendation algorithm that incorporates both intra and inter-metapath analysis, aiming to improve the model's understanding of complex linkages and dependencies between movies, users, and other entities.
Article
Computer Science, Information Systems
Bilal Ahmed, Li Wang, Ghulam Mustafa, Muhammad Tanvir Afzal, Adnan Akhunzada
Summary: Assessing the academic influence of researchers is a challenging task, and there is currently no universally accepted standard. This study computed 14 metrics to determine potential measures of influence and found high correlations among these metrics. Some metrics showed significant differences in rankings. Certain metrics were closely associated with award winners, and there was some relationship between specific societies and metrics.
Article
Computer Science, Theory & Methods
Sheng Wang, Shiping Chen, Fei Meng, Yumei Shi
Summary: This study proposes a Multi-Scenarios Adaptive Hierarchical Spatial Graph Convolution Network (MSHGN) model for accurately predicting GPU utilization rates in heterogeneous GPU clusters. By constructing multiple scenarios' undirected graphs and using Graph Convolution Neural (GCN) to capture spatial dependency relationships, the MSHGN model achieves superior accuracy and robustness in predicting resource utilization on a real-world Alibaba dataset.
JOURNAL OF PARALLEL AND DISTRIBUTED COMPUTING
(2024)
Article
Computer Science, Theory & Methods
Nik Amir Syafiq, Mohamed Othman, Norazak Senu, Fudziah Ismail, Nor Asilah Wati Abdul Hamid
Summary: This research investigates the multi-core architecture for solving the fractional Poisson equation using the modified accelerated overrelaxation (MAOR) scheme. The feasibility of the scheme in a parallel environment was tested through experimental comparisons and measurements. The results showed that the scheme is viable in a parallel environment.
JOURNAL OF PARALLEL AND DISTRIBUTED COMPUTING
(2024)
Article
Computer Science, Theory & Methods
Antonio F. Diaz, Beatriz Prieto, Juan Jose Escobar, Thomas Lampert
Summary: This paper presents the design and implementation of a low-cost energy monitoring system that synchronously collects the energy consumption of multiple devices using a specially designed wattmeter, and utilizes widely used technologies and tools in the Internet of Things for implementation.
JOURNAL OF PARALLEL AND DISTRIBUTED COMPUTING
(2024)
Article
Computer Science, Theory & Methods
Ritam Ganguly, Yingjie Xue, Aaron Jonckheere, Parker Ljung, Benjamin Schornstein, Borzoo Bonakdarpour, Maurice Herlihy
Summary: This paper presents a centralized runtime monitoring technique for distributed systems, which verifies the correctness of distributed computations by exploiting bounded-skew clock synchronization. By introducing a progression-based formula rewriting scheme and utilizing SMT solving techniques, the metric temporal logic can be monitored and the probabilistic guarantee for verification results can be calculated. Experimental results demonstrate the effectiveness of this technique in different application scenarios.
JOURNAL OF PARALLEL AND DISTRIBUTED COMPUTING
(2024)
Article
Computer Science, Theory & Methods
Arya Tanmay Gupta, Sandeep S. Kulkarni
Summary: Lattice-linear systems allow nodes to execute asynchronously. The eventually lattice-linear algorithms introduced in this study guarantee system transitions to optimal states within specified moves, leading to improved performance compared to existing literature. Experimental results further support the benefits of lattice-linearity.
JOURNAL OF PARALLEL AND DISTRIBUTED COMPUTING
(2024)
Article
Computer Science, Theory & Methods
Tim Breitenbach, Shrikanth Malavalli Divakar, Lauritz Rasbach, Patrick Jahnke
Summary: With the trend towards multi-socket server systems, the demand for RAM per server has increased, resulting in more DIMM sockets per server. RAM issues have become a dominant failure pattern for servers due to the probability of failure in each DIMM. This study introduces an ML-driven framework to estimate the probability of memory failure for each RAM module. The framework utilizes structural information between correctable (CE) and uncorrectable errors (UE) and engineering measures to mitigate the impact of UE.
JOURNAL OF PARALLEL AND DISTRIBUTED COMPUTING
(2024)
Article
Computer Science, Theory & Methods
Carlos Ansotegui, Eduard Torres
Summary: This paper presents an incomplete algorithm for efficiently constructing Covering Arrays with Constraints of high strength. The algorithm mitigates memory blow-ups and reduces run-time consumption, providing a practical tool for Combinatorial Testing.
JOURNAL OF PARALLEL AND DISTRIBUTED COMPUTING
(2024)
Article
Computer Science, Theory & Methods
Lucas Perotin, Sandhya Kandaswamy, Hongyang Sun, Padma Raghavan
Summary: Resource scheduling is crucial in High-Performance Computing systems, and previous research has mainly focused on a single type of resource. With advancements in hardware and the rise of data-intensive applications, considering multiple resources simultaneously is necessary to improve overall application performance. This study presents a Multi-Resource Scheduling Algorithm (MRSA) that minimizes the makespan of computational workflows by efficiently allocating resources and optimizing scheduling order. Simulation results demonstrate that MRSA outperforms baseline methods in various scenarios.
JOURNAL OF PARALLEL AND DISTRIBUTED COMPUTING
(2024)
Article
Computer Science, Theory & Methods
Yue Li, Han Liu, Jianbo Gao, Jiashuo Zhang, Zhi Guan, Zhong Chen
Summary: The processing of block lifecycles is crucial to the efficiency of a blockchain. The FASTBLOCK framework, which introduces fine-grained concurrency, accelerates the execution and validation steps. It outperforms state-of-the-art solutions significantly in terms of performance.
JOURNAL OF PARALLEL AND DISTRIBUTED COMPUTING
(2024)
Article
Computer Science, Theory & Methods
Roberto Carrasco, Hector Ferrada, Cristobal A. Navarro, Nancy Hitschfeld
Summary: The experimental evaluation of GPU filters for computing the 2D convex hull shows significant performance improvement. The different point distributions have a noticeable impact on the results, with the greatest improvement seen in the case of uniform and normal distributions.
JOURNAL OF PARALLEL AND DISTRIBUTED COMPUTING
(2024)
Article
Computer Science, Theory & Methods
Keqin Li
Summary: In this paper, the authors study task scheduling with or without energy constraint in mobile edge computing. They propose heuristic algorithms to solve these problems and analyze them using the methods of communication unification, effective speed concept, and virtual task construction. The experimental results show that the performance of the heuristic algorithms is close to the optimal algorithm. This is the first paper in the literature to optimize the makespan of task scheduling with or without energy constraint in mobile edge computing with multiple cloud-assisted edge servers.
JOURNAL OF PARALLEL AND DISTRIBUTED COMPUTING
(2024)
Article
Computer Science, Theory & Methods
Hongliang Li, Hairui Zhao, Ting Sun, Xiang Li, Haixiao Xu, Keqin Li
Summary: This paper studies the problem of job placement in shared GPU clusters and proposes an opportunistic memory sharing model and algorithms to solve the problem. Extensive experiments on a GPU cluster validate the correctness and effectiveness of the proposed approach.
JOURNAL OF PARALLEL AND DISTRIBUTED COMPUTING
(2024)
Article
Computer Science, Theory & Methods
Lucas Ruchel, Edson Tavares de Camargo, Luiz Antonio Rodrigues, Rogerio C. Turchetti, Luciana Arantes, Elias Procopio Duarte Jr.
Summary: LHABcast is a leaderless hierarchical atomic broadcast algorithm that improves scalability by being fully decentralized and hierarchical. It uses local sequence numbers and timestamps to order messages and achieves significantly lower message count compared to an all-to-all strategy, both in fault-free and faulty scenarios.
JOURNAL OF PARALLEL AND DISTRIBUTED COMPUTING
(2024)
Article
Computer Science, Theory & Methods
Xiangyu Wu, Xuehui Du, Qiantao Yang, Na Wang, Wenjuan Wang
Summary: This paper proposes a new method to address the immutability issue of consortium blockchains by introducing a verifiable distributed chameleon hash (VDCH) function and a consensus protocol called CVTSS based on verifiable threshold signatures. The proposed method enhances the flexibility, fault tolerance, and redaction efficiency of consortium blockchains.
JOURNAL OF PARALLEL AND DISTRIBUTED COMPUTING
(2024)
Article
Computer Science, Theory & Methods
Ipsita Behera, Srichandan Sobhanayak
Summary: Task scheduling in cloud computing is a challenging problem, and researchers propose a hybrid algorithm that aims to minimize makespan, energy consumption, and cost. Evaluation using the Cloudsim toolkit demonstrates the algorithm's effectiveness and efficiency.
JOURNAL OF PARALLEL AND DISTRIBUTED COMPUTING
(2024)