Article
Computer Science, Information Systems
Ling Chen, Yi Zhang, Shenghuan Miao, Sirou Zhu, Rong Hu, Liangying Peng, Mingqi Lv
Summary: Unsupervised user adaptation aligns the feature distributions of training users and new user for wearable human activity recognition (WHAR) model adaptation. We propose SALIENCE model for multiple wearable sensors based WHAR, which addresses the challenge of different sensor transferabilities by separate local alignment and uniform global alignment. An attention mechanism is introduced to focus on sensors with strong feature discrimination and well distribution alignment. Experimental results show competitive performance.
IEEE TRANSACTIONS ON MOBILE COMPUTING
(2023)
Article
Computer Science, Artificial Intelligence
Jinhyeok Kim, Jongsoo Lee
Summary: The availability of high-quality data is crucial for the performance of machine learning models. This paper proposes an instance-based transfer learning method to overcome data scarcity and improve model accuracy by reusing data from similar models. The results of three case studies indicate a significant improvement in neural network prediction despite data scarcity.
APPLIED SOFT COMPUTING
(2022)
Article
Computer Science, Information Systems
Yida Zhu, Haiyong Luo, Song Guo, Fang Zhao
Summary: Human activity recognition (HAR) based on wearable sensors is a popular research topic. Obtaining labeled human activity data for different body-worn positions is expensive and labor-intensive, leading to poor performance of HAR models on different body positions. In this article, we propose a deep multiscale transfer learning (DMSTL) model for accurate HAR with low labeling cost. The model includes an unsupervised source selection method, a multiscale spatial-temporal Net (MSSTNet) for comprehensive multimodal representations, and category-level adaptation and domain-level adversarial modules for learning domain-invariant features. Experimental results on three public HAR datasets show that DMSTL outperforms other baselines.
IEEE INTERNET OF THINGS JOURNAL
(2023)
Article
Chemistry, Analytical
Tsige Tadesse Alemayoh, Jae Hoon Lee, Shingo Okamoto
Summary: This study introduces a new method for activity recognition using smartphone data collection and deep learning classification with neural network models. The experimental results show good performance, with better performance validated on other datasets, and the practicality of the model was demonstrated in real-time testing on a computer and smartphone.
Article
Forestry
Zhengjun Yan, Liming Wang, Kui Qin, Feng Zhou, Jineng Ouyang, Teng Wang, Xinguo Hou, Leping Bu
Summary: This paper applies unsupervised domain adaptation (UDA) to transfer knowledge from a labeled public fire dataset to another unlabeled one, providing a benchmark dataset for fire recognition. Two adaptation networks are experimented on this dataset and achieve impressive performance in fire recognition.
Article
Computer Science, Information Systems
Quanle Liu, Xiangjiu Che, Mengxue Zhao
Summary: The wide application of vision sensors has created both opportunities and challenges for intelligent understanding of image and video data. Efficient processing and utilization of this data necessitates research on effective algorithms for human activity understanding. Existing methods for human activity understanding suffer from issues such as low recognition accuracy, high complexity, and poor robustness. In response to these problems, the body part relation reasoning network (BPRRN) is constructed to leverage common-sense knowledge within human bodies for reasoning human activities. Experimental results on Stanford 40 and UCF101 datasets demonstrate the effectiveness of the proposed network.
INFORMATION SCIENCES
(2023)
Article
Telecommunications
Jayita Saha, Dip Ghosh, Chandreyee Chowdhury, Sanghamitra Bandyopadhyay
Summary: The study designed a human activity recognition framework that can identify multiple activities, leveraging the temporal relationship among activities to obtain a more comprehensive model, which was found to work adequately during testing.
WIRELESS PERSONAL COMMUNICATIONS
(2021)
Article
Chemistry, Analytical
Jan Slemensek, Iztok Fister, Jelka Gersak, Bozidar Bratina, Vesna Marija van Midden, Zvezdan Pirtosek, Riko Safaric
Summary: This paper proposes a robust, wearable gait motion data acquisition system that allows the classification of recorded gait data into desirable activities and the identification of common risk factors, thus enhancing the subject's quality of life. Gait motion information was acquired using accelerometers and gyroscopes mounted on the lower limbs, and leg muscle activity was measured using strain gauge sensors. Machine Learning algorithms were utilized to identify different gait activities within each gait recording, with the combination of attention-based convolutional and recurrent neural networks algorithms outperforming others.
Article
Automation & Control Systems
Yinjun Wang, Liang Ge, Chunrong Xue, Xiaobo Li, Xianghui Meng, Xiaoxi Ding
Summary: Deep learning has been widely used in mechanical fault diagnosis and equipment health monitoring. A practical issue is the cross-domain machinery fault diagnosis, where the models are typically trained with artificial fault samples due to limited real fault samples. However, the characteristics of artificial fault samples in the lab differ from real fault samples in the industrial environment. This study proposes a multiple local domains transfer network to address this issue by reducing negative transfer through multi-local domain adversarial learning.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2023)
Article
Computer Science, Artificial Intelligence
Aria Ghora Prabono, Bernardo Nugroho Yahya, Seok-Lyong Lee
Summary: This study introduces a domain-invariant latent representation to enhance specific feature transfer in the hybrid domain adaptation approach. The learning of latent representation and domain-specific feature transfer are performed simultaneously using an autoencoder-based framework. Experimental results show that performance improves when common features are further aligned in the latent space.
PATTERN ANALYSIS AND APPLICATIONS
(2021)
Article
Geochemistry & Geophysics
Jin Wang, Shunping Ji, Tao Zhang
Summary: Remote sensing images can vary significantly due to factors like atmospheric conditions and sensor types, making it challenging to apply pretrained instance segmentation models to new images. Existing domain adaptation methods are not tailored for cross-domain instance segmentation, which requires aligning object-level features instead of whole images. To address this, we propose a method based on object-level alignment, incorporating an improved contour-based instance segmentation model, a Fourier domain adaptation technique for object-pasting enhancement, and a self-training strategy. Experimental results on different datasets demonstrate the effectiveness and universality of our approach, achieving substantial improvements in intersection over union (IoU) and mean average precision (mAP) compared to current state-of-the-art methods.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Agriculture, Dairy & Animal Science
Seong-Ho Ahn, Seeun Kim, Dong-Hwa Jeong
Summary: This study aimed to improve animal activity recognition (AAR) using wearable sensor data by addressing challenges such as sensor and individual variability. The use of unsupervised domain adaptation (UDA) techniques significantly improved classification performance by mitigating these variabilities. The findings highlight the practical applications of UDA in real-world scenarios with limited labeled data.
Article
Geochemistry & Geophysics
Weibin Song, Xuping Feng, Gongheng Zhang, Lina Gao, Binpeng Yan, Xiaofei Chen
Summary: This study applies ambient seismic noise tomography to probe the Earth's structure and successfully enhances the generalization of neural networks for better extraction of dispersion information through the introduction of domain adaptation method.
JOURNAL OF GEOPHYSICAL RESEARCH-SOLID EARTH
(2022)
Article
Engineering, Electrical & Electronic
Zhongping Cao, Zhenchang Li, Xuemei Guo, Guoli Wang
Summary: This paper addresses the issue of adapting radar-based human activity recognition systems to new environments without source data. By utilizing the source hypothesis transfer learning architecture, a mechanism for cross-environment adaptation of radar-based HAR is developed. A reliable self-supervised labeling strategy is proposed to generate pseudo labels for unlabeled target data, leading to the improvement of target-specific feature extraction for environment adaptation. The experimental results demonstrate the effectiveness of the proposed approach on a public HAR dataset.
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY
(2021)
Article
Engineering, Multidisciplinary
Yaochun Wu, Rongzhen Zhao, Hongru Ma, Qiang He, Shaohua Du, Jie Wu
Summary: This study proposes an intelligent recognition method based on ADACNN to overcome the limitation of traditional methods in bearing fault detection caused by variations in working conditions.
Editorial Material
Computer Science, Theory & Methods
Kiho Lim, Christian Esposito, Tian Wang, Chang Choi
FUTURE GENERATION COMPUTER SYSTEMS-THE INTERNATIONAL JOURNAL OF ESCIENCE
(2024)
Editorial Material
Computer Science, Theory & Methods
Jesus Carretero, Dagmar Krefting
Summary: Computational methods play a crucial role in bioinformatics and biomedicine, especially in managing large-scale data and simulating complex models. This special issue focuses on security and performance aspects in infrastructure, optimization for popular applications, and the integration of machine learning and data processing platforms to improve the efficiency and accuracy of bioinformatics.
FUTURE GENERATION COMPUTER SYSTEMS-THE INTERNATIONAL JOURNAL OF ESCIENCE
(2024)
Article
Computer Science, Theory & Methods
Renhao Lu, Weizhe Zhang, Qiong Li, Hui He, Xiaoxiong Zhong, Hongwei Yang, Desheng Wang, Zenglin Xu, Mamoun Alazab
Summary: Federated Learning allows collaborative training of AI models with local data, and our proposed FedAAM scheme improves convergence speed and training efficiency through an adaptive weight allocation strategy and asynchronous global update rules.
FUTURE GENERATION COMPUTER SYSTEMS-THE INTERNATIONAL JOURNAL OF ESCIENCE
(2024)
Article
Computer Science, Theory & Methods
Qiangqiang Jiang, Xu Xin, Libo Yao, Bo Chen
Summary: This paper proposes a multi-objective energy-efficient task scheduling technique (METSM) for edge heterogeneous multiprocessor systems. A mathematical model is established for the task scheduling problem, and a problem-specific algorithm (IMO) is designed for optimizing task scheduling and resource allocation. Experimental results show that the proposed algorithm can achieve optimal Pareto fronts and significantly save time and power consumption.
FUTURE GENERATION COMPUTER SYSTEMS-THE INTERNATIONAL JOURNAL OF ESCIENCE
(2024)
Editorial Material
Computer Science, Theory & Methods
Weimin Li, Lu Liu, Kevin I. K. Wang, Qun Jin
FUTURE GENERATION COMPUTER SYSTEMS-THE INTERNATIONAL JOURNAL OF ESCIENCE
(2024)
Article
Computer Science, Theory & Methods
Mohammed Riyadh Abdmeziem, Amina Ahmed Nacer, Nawfel Moundji Deroues
Summary: Internet of Things (IoT) devices have become ubiquitous and brought the need for group communications. However, security in group communications is challenging due to the asynchronous nature of IoT devices. This paper introduces an innovative approach using blockchain technology and smart contracts to ensure secure and scalable group communications.
FUTURE GENERATION COMPUTER SYSTEMS-THE INTERNATIONAL JOURNAL OF ESCIENCE
(2024)
Article
Computer Science, Theory & Methods
Robert Sajina, Nikola Tankovic, Ivo Ipsic
Summary: This paper presents and evaluates a novel approach that utilizes an encoder-only transformer model to enable collaboration between agents learning two distinct NLP tasks. The evaluation results demonstrate that collaboration among agents, even when working towards separate objectives, can result in mutual benefits.
FUTURE GENERATION COMPUTER SYSTEMS-THE INTERNATIONAL JOURNAL OF ESCIENCE
(2024)
Article
Computer Science, Theory & Methods
Hebert Cabane, Kleinner Farias
Summary: Event-driven architecture has been widely adopted in the software industry for its benefits in software modularity and performance. However, there is a lack of empirical evidence to support its impact on performance. This study compares the performance of an event-driven application with a monolithic application and finds that the monolithic architecture consumes fewer computational resources and has better response times.
FUTURE GENERATION COMPUTER SYSTEMS-THE INTERNATIONAL JOURNAL OF ESCIENCE
(2024)
Article
Computer Science, Theory & Methods
Haroon Wahab, Irfan Mehmood, Hassan Ugail, Javier Del Ser, Khan Muhammad
Summary: Wireless capsule endoscopy (WCE) is a revolutionary diagnostic method for small bowel pathology. However, the manual analysis of WCE videos is cumbersome and the privacy concerns of WCE data hinder the adoption of AI-based diagnoses. This study proposes a federated learning framework for collaborative learning from multiple data centers, demonstrating improved anomaly classification performance while preserving data privacy.
FUTURE GENERATION COMPUTER SYSTEMS-THE INTERNATIONAL JOURNAL OF ESCIENCE
(2024)
Article
Computer Science, Theory & Methods
Maruf Monem, Md Tamjid Hossain, Md. Golam Rabiul Alam, Md. Shirajum Munir, Md. Mahbubur Rahman, Salman A. AlQahtani, Samah Almutlaq, Mohammad Mehedi Hassan
Summary: Bitcoin, the largest cryptocurrency, faces challenges in broader adaption due to long verification times and high transaction fees. To tackle these issues, researchers propose a learning framework that uses machine learning to predict the ideal block size in each block generation cycle. This model significantly improves the block size, transaction fees, and transaction approval rate of Bitcoin, addressing the long wait time and broader adaption problem.
FUTURE GENERATION COMPUTER SYSTEMS-THE INTERNATIONAL JOURNAL OF ESCIENCE
(2024)
Article
Computer Science, Theory & Methods
Rafael Duque, Crescencio Bravo, Santos Bringas, Daniel Postigo
Summary: This paper introduces the importance of user interfaces for digital twins and presents a technique called ADD for modeling requirements of Human-DT interaction. A study is conducted to assess the feasibility and utility of ADD in designing user interfaces, using the virtualization of a natural space as a case study.
FUTURE GENERATION COMPUTER SYSTEMS-THE INTERNATIONAL JOURNAL OF ESCIENCE
(2024)
Article
Computer Science, Theory & Methods
Xiulin Li, Li Pan, Wei Song, Shijun Liu, Xiangxu Meng
Summary: This article proposes a novel multiclass multi-pool analytical model for optimizing the quality of composite service applications deployed in the cloud. By considering embarrassingly parallel services and using differentiated parallel processing mechanisms, the model provides accurate prediction results and significantly reduces job response time.
FUTURE GENERATION COMPUTER SYSTEMS-THE INTERNATIONAL JOURNAL OF ESCIENCE
(2024)
Article
Computer Science, Theory & Methods
Seongwan Park, Woojin Jeong, Yunyoung Lee, Bumho Son, Huisu Jang, Jaewook Lee
Summary: In this paper, a novel MEV detection model called ArbiNet is proposed, which offers a low-cost and accurate solution for MEV detection without requiring knowledge of smart contract code or ABIs.
FUTURE GENERATION COMPUTER SYSTEMS-THE INTERNATIONAL JOURNAL OF ESCIENCE
(2024)
Article
Computer Science, Theory & Methods
Sacheendra Talluri, Nikolas Herbst, Cristina Abad, Tiziano De Matteis, Alexandru Iosup
Summary: Serverless computing is increasingly used in data-processing applications. This paper presents ExDe, a framework for systematically exploring the design space of scheduling architectures and mechanisms, to help system designers tackle complexity.
FUTURE GENERATION COMPUTER SYSTEMS-THE INTERNATIONAL JOURNAL OF ESCIENCE
(2024)
Article
Computer Science, Theory & Methods
Chao Wang, Hui Xia, Shuo Xu, Hao Chi, Rui Zhang, Chunqiang Hu
Summary: This paper introduces a Federated Learning framework called FedBnR to address the issue of potential data heterogeneity in distributed entities. By breaking up the original task into multiple subtasks and reconstructing the representation using feature extractors, the framework improves the learning performance on heterogeneous datasets.
FUTURE GENERATION COMPUTER SYSTEMS-THE INTERNATIONAL JOURNAL OF ESCIENCE
(2024)