Article
Computer Science, Interdisciplinary Applications
Pilar Lopez-Ubeda, Manuel Carlos Diaz-Galiano, Teodoro Martin-Noguerol, Antonio Luna, L. Alfonso Urena-Lopez, M. Teresa Martin-Valdivia
Summary: This study developed machine learning classification models based on NLP using patient data from radiological reports and protocols. The best proposed system achieved 92.2% accuracy in the CT dataset and 86.9% in the MRI dataset. The potential efficiency, quality, and cost-effectiveness make the best machine learning system currently used in real scenarios by radiologists as a decision support tool for protocol assignment in CT and MRI studies.
COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE
(2021)
Article
Engineering, Electrical & Electronic
Rencan Nie, Jinde Cao, Dongming Zhou, Wenhua Qian
Summary: This paper proposes a novel fusion framework for multimodal medical images based on PCNN and MIEE. The method exchanges and encodes information between different images, producing quantitative fusion contributions. Experimental results show that this approach outperforms other state-of-the-art fusion methods.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2021)
Article
Computer Science, Information Systems
Liangyi Kang, Jie Liu, Lingqiao Liu, Zhiyang Zhou, Dan Ye
Summary: This paper introduces a semi-supervised ERC algorithm that leverages unlabeled conversational data with a novel Context-augmented AUXIliary training Task (CAUXIT) to improve ERC model performance. By predicting emotion-related information of utterances based on context, the network's ability in making emotion inference is enhanced, resulting in significant improvement over traditional supervised ERC methods.
INFORMATION PROCESSING & MANAGEMENT
(2021)
Article
Biology
Shaoxiong Ji, Matti Holtta, Pekka Marttinen
Summary: Unsupervised pretraining is essential in natural language processing, and this paper analyzes the performances of various contextualized language models for medical code assignment from clinical notes. A hierarchical fine-tuning architecture and label-wise attention mechanism are proposed in this study. Contrary to current trends, a carefully trained CNN outperforms attention-based models on a specific dataset, suggesting directions for building robust medical code assignment models.
COMPUTERS IN BIOLOGY AND MEDICINE
(2021)
Article
Computer Science, Hardware & Architecture
Shiyi Qi, Yaoxian Li, Cuiyun Gao, Xiaohong Su, Shuzheng Gao, Zibin Zheng, Chuanyi Liu
Summary: This study proposes a Transformer-based model, named DTrans, for learning to predict code changes. By incorporating dynamically relative position encoding in the multi-head attention of Transformer, DTrans can accurately generate patches and locate lines to change with higher accuracy compared to existing methods.
IEEE TRANSACTIONS ON RELIABILITY
(2023)
Article
Engineering, Electrical & Electronic
Yutao Hu, Yandan Yang, Jun Zhang, Xianbin Cao, Xiantong Zhen
Summary: The study proposes an Attentional Kernel Encoding Networks (AKEN) for fine-grained visual categorization, which aggregates and encodes feature maps, and incorporates a Cascaded Attention module for discriminative feature extraction. Compared to traditional methods, AKEN shows highly competitive performance in fine-grained image categorization tasks.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2021)
Article
Computer Science, Artificial Intelligence
Suman Dowlagar, Radhika Mamidi
Summary: In the healthcare domain, the interactions between medical professionals and patients are crucial for diagnosis. However, existing AI models in healthcare are designed for monolingual data and cannot handle code-mixed conversations in multilingual regions. To facilitate the research and development of code-mixed medical dialog systems, we introduce a dataset of code-mixed medical dialogs and provide baselines for benchmarking.
COMPUTER SPEECH AND LANGUAGE
(2023)
Article
Genetics & Heredity
Jacques Demongeot, Herve Seligmann
Summary: The genetic code codon-amino acid assignments evolve within 15 out of 64 codons (23.4%) across 31 genetic codes. The use of evolvable codons in RNA rings is positively correlated with recent cognates and cytosine numbers. Alternative genetic codes applied to RNA rings designed for nonredundant coding according to the standard genetic code reveal new insights on the evolvability of codon assignments and the role of cytosine in genetic code evolution.
Article
Computer Science, Artificial Intelligence
Alice Witt, Anna Huggins, Guido Governatori, Joshua Buckley
Summary: This article proposes an innovative methodology to enhance the technical validation, legal alignment, and interdisciplinarity of encoding legislation. Through an experiment, the study finds that a combination of manual and automated methods for coding validation can significantly increase the similarity of encoded rules between coders. However, legal evaluation is necessary to address interpretive difficulties, and a process of "legal alignment" can enhance the congruence between encoded provisions and the true meaning of a statute.
ARTIFICIAL INTELLIGENCE AND LAW
(2023)
Article
Computer Science, Artificial Intelligence
Changan Niu, Chuanyi Li, Vincent Ng, Bin Luo
Summary: Recent years have witnessed the successful application of large pretrained models of source code (CodePTMs) in code representation learning, transforming the field of software engineering from task-specific solutions to generic models. However, the lack of comparability among CodePTMs due to differences in experimental setups has posed a challenge. This article proposes a standardized setup to enable fair comparisons and explores the impact of pretraining tasks on CodePTMs, presenting experimental results and comprehensive discussions to advance the future study of more powerful CodePTMs.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Engineering, Electrical & Electronic
Kaiyou Song, Hua Yang, Zhouping Yin
Summary: The study proposes a novel multi-scale boosting feature encoding network (MSBFEN) for accurate texture recognition by extracting multi-scale features using texture priors and then encoding them, along with a multi-scale boosting learning (MSBL) method to enhance recognition accuracy. Results show state-of-the-art performance on challenging texture recognition datasets.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2021)
Article
Computer Science, Information Systems
Hongjun Xu, Narushan Pillay
Summary: This paper proposes an alternative encoding method for the Golden code with simplified and low complexity implementation. An equivalent received signal model is constructed to derive the lower bound of average bit error probability. Additionally, a low complexity detection scheme is proposed to reduce detection complexity at high signal-to-noise ratios.
Article
Computer Science, Information Systems
Nan Chen, Xiaoyu Ma, Yong Chen, Dingguo Yu
Summary: Images and short videos produced by social networks have experienced a surge in recent years, requiring the use of image/video encoders to reduce transmission bandwidth. However, in real-world scenarios, encoding parameters are often preset to fixed values, which may not be the optimal bandwidth allocation strategy. Therefore, an efficient group quality optimization framework has been proposed to optimize encoding parameters based on perceptual quality.
Article
Computer Science, Artificial Intelligence
Philip Easom-McCaldin, Ahmed Bouridane, Ammar Belatreche, Richard Jiang, Somaya Al-Maadeed
Summary: This article proposes a new single-qubit-based deep quantum neural network for image classification, which mimics traditional convolutional neural network techniques and reduces the number of parameters. Preliminary test results on various datasets demonstrate promising classification accuracies.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Computer Science, Software Engineering
Aakash Bansal, Zachary Eberhart, Zachary Karas, Yu Huang, Collin Mcmillan
Summary: Source code summarization is the task of writing natural language descriptions of source code, primarily used for programmer documentation. Automatic generation of these descriptions is a valuable research goal to save programmers' time. However, most approaches tend to rely solely on the context of the source code being summarized, while empirical studies suggest that the necessary information to describe the code often resides in the context, such as the Function Call Graph. This paper presents a technique for encoding this call graph context for neural models of code summarization, showing statistically significant improvement over existing approaches and perceived accuracy, readability, and conciseness in human evaluation.
IEEE TRANSACTIONS ON SOFTWARE ENGINEERING
(2023)
Article
Computer Science, Hardware & Architecture
Fei Teng, Dian Wang, Yue Yuan, Haibo Zhang, Amit Kumar Singh, Zhihan Lv
Summary: This study proposes a deep active learning model to detect OSA events from electrocardiogram (ECG), and develops a prototype of OSA monitoring system using ECG sensor and smartphone. The system provides OSA risk level and medical advice based on detection results and health-related multimedia signals.
Article
Engineering, Civil
Penglin Dai, Kaiwen Hu, Xiao Wu, Huanlai Xing, Fei Teng, Zhaofei Yu
Summary: This article investigates the computation offloading problem in MEC-assisted vehicular networks, considering task upload coordination, task migration, and heterogeneous computation capabilities of MEC/cloud servers. A probabilistic computation offloading (PCO) algorithm is proposed to minimize task completion delay based on queuing theory. Simulation results demonstrate the superiority of the proposed algorithm in various scenarios.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2022)
Article
Computer Science, Information Systems
Jie Yin, Yang Xiao, Qingqi Pei, Ying Ju, Lei Liu, Ming Xiao, Celimuge Wu
Summary: In this article, we propose SmartDID, a novel blockchain-based distributed identity that aims to establish a self-sovereign identity and provide strong privacy preservation. Our method configures IoT devices as light nodes and designs a Sybil-resistant, unlinkable, and supervisable distributed identity. We further develop a dual-credential model based on commitment and zero-knowledge proofs to protect privacy. Our scheme achieves better performance compared to CanDID in terms of both credential generation and proof generation.
IEEE INTERNET OF THINGS JOURNAL
(2023)
Article
Computer Science, Information Systems
Hong Niu, Xia Lei, Yue Xiao, Ming Xiao, Shahid Mumtaz
Summary: In order to improve physical-layer security, reconfigurable intelligent surface (RIS) has been proposed to assist artificial noise (AN) in MISO systems. By jointly designing phase shifts (PSs) and beamforming based on the oblique manifold (OM) algorithm, the trade-off between transmission reliability and security is achieved. The novel alternating direction (AD) algorithm is introduced to lower computational complexity and exhibits faster and more stable convergence compared to the OM algorithm.
IEEE WIRELESS COMMUNICATIONS LETTERS
(2022)
Review
Physics, Multidisciplinary
Ming Xiao, Mikael Skoglund
Summary: This article provides a comprehensive and rigorous review of the principles and recent development of coding for large-scale distributed machine learning (DML), aiming to improve reliability and efficiency. Various coding schemes for different steps in DML are discussed, with potential directions for future works also provided.
Article
Engineering, Electrical & Electronic
Yixuan Huang, Su Hu, Shiyong Ma, Zilong Liu, Ming Xiao
Summary: This paper proposes a waveform design algorithm for reducing PAPR in RadCom systems. By optimizing subcarriers, it achieves significant performance enhancements compared to legacy approaches.
IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS
(2022)
Article
Engineering, Electrical & Electronic
Sun Mao, Kun Yang, Jie Hu, Dong Li, Dapeng Lan, Ming Xiao, Youzhi Xiong
Summary: Wireless powered hybrid backscatter-active communications are studied in this paper. Multiple sensor devices, powered by a power beacon, transmit their information to an access point using backscatter and active communications with the assistance of an intelligent reflecting surface. The proposed method optimizes the system energy efficiency by jointly optimizing the phase shifts of the intelligent reflecting surface, time allocation for communications, and transmit power of the devices. Simulation results show that the proposed method outperforms existing benchmark methods in terms of system energy efficiency.
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY
(2023)
Article
Engineering, Civil
Ying Ju, Yuchao Chen, Zhiwei Cao, Lei Liu, Qingqi Pei, Ming Xiao, Kaoru Ota, Mianxiong Dong, Victor C. M. Leung
Summary: This paper proposes a deep reinforcement learning based joint secure offloading and resource allocation scheme to improve the secrecy performance and resource efficiency of multi-user vehicular edge computing networks.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2023)
Article
Computer Science, Information Systems
Yiqian Huang, Ping Yang, Bo Zhang, Zilong Liu, Ming Xiao
Summary: The vision of 6G is to achieve reliable data transmission in complex and diverse wireless channels. A new paradigm is proposed by combining MDS codes with OFDM and utilizing RIS. The proposed system utilizes RIS with limited phase shifts and a novel method based on phase alignment to minimize BER. Simulation results demonstrate significant BER improvements compared to conventional OFDM systems and similar performance to systems with continuous phase shifts is achieved by the proposed optimization algorithm.
IEEE WIRELESS COMMUNICATIONS LETTERS
(2023)
Article
Engineering, Electrical & Electronic
Zijun Qin, Zesong Fei, Jingxuan Huang, Yeliang Wang, Ming Xiao, Jinhong Yuan
Summary: In this paper, we investigate the application of reinforcement learning (RL) in online fountain codes and propose two schemes to reduce the overhead of full recovery with limited feedback. The first scheme, RL-based degree determination (RL-DD), uses RL to determine the optimal degree of coded symbols based on theoretical analysis. The second scheme, online fountain codes with no build-up phase using sectioned distribution (OFCNB-SD), eliminates the build-up phase and improves the decoder to utilize non-immediately decodable coded symbols. The simulation results demonstrate that our proposed schemes achieve lower full-recovery overhead compared to existing schemes.
IEEE TRANSACTIONS ON COMMUNICATIONS
(2023)
Article
Computer Science, Information Systems
Nan Qi, Zanqi Huang, Fuhui Zhou, Qingjiang Shi, Qihui Wu, Ming Xiao
Summary: A sequential overlapping coalition formation game model is proposed to optimize the composition and task allocation of heterogeneous UAVs. The model considers the overlapping and complementary relations of resource properties and the task execution order, and introduces a bilateral mutual benefit transfer order to allocate task resources more efficiently.
IEEE TRANSACTIONS ON MOBILE COMPUTING
(2023)
Article
Computer Science, Information Systems
Teng Ma, Yue Xiao, Xia Lei, Wenhui Xiong, Ming Xiao
Summary: The article introduces the reconfigurable intelligent surface (RIS) as a revolutionary technology for future wireless communications. It has been widely studied in recent years for enhancing signal quality, energy efficiency, and throughput. However, in RIS-assisted IoT, a new issue of multipath separation arises when deploying multiple RISs. To alleviate this problem, a novel RIS-enabled code-division multiple access (CDMA) structure is proposed, along with practical channel estimation and theoretical derivations. Simulation results confirm the feasibility of the conceived RIS-CDMA structure and the effectiveness of the proposed multipath extraction approach.
IEEE INTERNET OF THINGS JOURNAL
(2023)
Article
Computer Science, Information Systems
Yanping Liu, Xuming Fang, Ming Xiao, Yaping Cui, Qing Xue
Summary: This article investigates how to overcome the problem of transmission blockage in millimeter-wave communication by coordinating multi-beam transmission and selecting appropriate transmission power to improve the system's stability and reliability. By constructing a hierarchical game model and designing a decentralized algorithm, the problem can be effectively solved.
IEEE WIRELESS COMMUNICATIONS LETTERS
(2023)
Article
Computer Science, Information Systems
Hao Chen, Yu Ye, Ming Xiao, Mikael Skoglund
Summary: This paper investigates the application of distributed learning in machine learning to alleviate the burden on central servers. Incremental block-coordinate descent (I-BCD) and asynchronous parallel incremental BCD (API-BCD) methods are proposed to reduce communication costs and accelerate convergence speed.
IEEE TRANSACTIONS ON BIG DATA
(2023)
Article
Computer Science, Information Systems
Kewei Wang, Nan Qi, Xin Guan, Qingjiang Shi, Ming Xiao, Shi Jin, Kai-Kit Wong
Summary: This article proposes a transmit/passive beamforming strategy in a multi-IRS assisted cell-free multi-input-multi-output network to maximize the weighted sum-rate. The elementwise block coordinate descent framework is applied for efficient solution of unit-modulus constraints and reduced computation complexity. Experimental results demonstrate that the proposed strategy outperforms benchmark algorithms with closed-form solutions in terms of complexity cost, while guaranteeing performance.
IEEE SYSTEMS JOURNAL
(2023)