A5100415541- Topic Modeling
- Natural Language Processing Techniques
- Vehicular Ad Hoc Networks (VANETs)
- Advanced Text Analysis Techniques
- Autonomous Vehicle Technology and Safety
- Multimodal Machine Learning Applications
- Advanced Malware Detection Techniques
- Cryptographic Implementations and Security
- Semantic Web and Ontologies
- Network Security and Intrusion Detection
- Advanced Memory and Neural Computing
- Domain Adaptation and Few-Shot Learning
- Advanced Database Systems and Queries
- Data Quality and Management
- Machine Learning and ELM
- IoT and GPS-based Vehicle Safety Systems
- Nonlinear Optical Materials Studies
- Advanced Technology in Applications
- Digital Media Forensic Detection
- Advanced Neural Network Applications
- Isotope Analysis in Ecology
- Network Traffic and Congestion Control
- E-commerce and Technology Innovations
- Adversarial Robustness in Machine Learning
- Handwritten Text Recognition Techniques
Soochow University
2024
Tianjin University of Science and Technology
2024
China Automotive Technology and Research Center
2017-2023
Harbin University of Science and Technology
2022
China University of Mining and Technology
2015
Nanjing University
2013-2014
Curtin University
2004
Timeline generation is of great significance for a comprehensive understanding the development events over time. Its goal to organize news chronologically, which helps identify patterns and trends that may be obscured when viewing in isolation, making it easier track stories understand interrelationships between key events. Timelines are now common various commercial products, but academic research this area notably scarce. Additionally, current datasets need refinement enhanced utility...
Extreme learning machine is a new algorithm for the single hidden layer feedforward neural networks (SLFNs). ELM has been widely used in various fields and applications to overcome slow training speed over-fitting problems of conventional network algorithms. based on empirical risk minimization, without considering structural this may lead at same time, it with poor controllability robustness. For these deficiencies, an optimization method proposed paper, novel extreme hybrid kernel function...
Joint entity and relation extraction (RE) construct a framework for unifying recognition relationship extraction, the approach can exploit dependencies between two tasks to improve performance of task. However, existing still have following problems. First, when model extracts information, boundary is blurred. Secondly, there are mostly implicit interactions modules, that is, interactive information hidden inside model, often insufficient in degree interaction lack interpretability. To this...
Developing single-pixel full-color liquid crystal displays (LCDs) that do not require orientation layers and color filters is highly desirable since this would allow to better optimize their image resolution light utilization efficiency while considerably reducing fabrication cost. However, so far, organic polymers have shown only limited modulation range inorganic materials mostly been on-and-off switches. Here, we report along with gray scale control in electrically responsive...
With the continuous development and upgrading of Intelligent Connected Vehicles (ICV), on-board network communication intelligent connected vehicles has been gradually transformed from traditional CAN bus to complex such as CAN-FD, FlexRay Ethernet. However, while architecture brings performance greatly improved, it also hidden cyber security risk. This paper mainly studies different buses perspective vehicle information security, proposes protection method Ethernet, carries on design...
Haolin Deng, Yanan Zhang, Yangfan Wangyang Ying, Changlong Yu, Jun Gao, Wei Wang, Xiaoling Bai, Nan Yang, Jin Ma, Xiang Chen, Tianhua Zhou. Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing. 2022.
To protect the confidentiality of information system, encryption techniques are essential and widely used. However, with a traditional block cipher may cause damage to application or database due format changing. Format-preserving (FPE) plays an important role in practice, especially for integer. We studied problems on applying FPE integer, specifically low performance caused by cycle-walking. In this paper, we suggest construction integer which is based unbalanced Feistel network. It...
With the rapid development of intelligent network connected vehicle, information security has been paid more and attention. In order to solve problem that vehicle bus is transmitted directly without encryption, which easy be illegally monitored maliciously cracked, it planned encrypt before transmission. Sender receiver use same encryption algorithm key. The sender encrypts data sending message, while decrypts after receiving it. study influence CAN decryption on transmission efficiency,...
The rotation symmetric Boolean functions which are invariant under the action of cyclic group have been used as components different cryptosystems. In order to resist algebraic attacks, should high immunity. This paper studies construction even-variable with optimum We construct
Though feature-alignment based Domain Adaptive Object Detection (DAOD) have achieved remarkable progress, they ignore the source bias issue, i.e. aligned features are more favorable towards domain, leading to a sub-optimal adaptation. Furthermore, presence of domain shift between and target domains exacerbates problem inconsistent classification localization in general detection pipelines. To overcome these challenges, we propose novel Distillation-based Unbiased Alignment (DUA) framework...
With the rise of pre-trained language models, few-shot learning has experienced significant progress in terms performance. Nevertheless, there remains considerable scope for improvement. The objective this study is to improve efficacy by introducing an enhanced prompt approach that maximizes utilization limited supervised information. Furthermore, our integrates a deep clustering method leverage unlabeled data, thereby further enhancing performance learning. Our ongoing experiments have...
Ensembles of generative large language models (LLMs) can integrate the strengths different LLMs to compensate for limitations individual models. However, recent work has focused on training an additional fusion model combine complete responses from multiple LLMs, failing tap into their collaborative potential generate higher-quality responses. Moreover, as is trained a specialized dataset, these methods struggle with generalizing open-domain queries online users. In this paper, we propose...
In order to address the shortcoming of feature representation limitation in machine translation(MT) system, this paper presents a transfer method MT. Meta decoding process considered not only their own translation but also transferred knowledge another system. The domain meta and objective function adaptation are used better model task. paper, extensive experiments comparisons made. experiment results show that proposed has significant improvement first performance than baseline which...
Event extraction (EE) is crucial to downstream tasks such as new aggregation and event knowledge graph construction. Most existing EE datasets manually define fixed types design specific schema for each of them, failing cover diverse events emerging from the online text. Moreover, news titles, an important source mentions, have not gained enough attention in current research. In this paper, We present Title2Event, a large-scale sentence-level dataset benchmarking Open Extraction without...
Abstract In the rapid development of intelligent network connected vehicle, many information security risks have been exposed. The application CAN bus encryption and decryption technology can solve problem safety communication in vehicle intranet. sender receiver use same set key. encrypts data before sending it, decrypts applying it. order to study influence on transmission efficiency, taking tc299, s32k144, mpc5606b spc560b54 chips as examples, different algorithms are used encrypt decrypt...