A5100347847- Natural Language Processing Techniques
- Topic Modeling
- Mobile Crowdsensing and Crowdsourcing
- Privacy-Preserving Technologies in Data
- Speech Recognition and Synthesis
- Vehicular Ad Hoc Networks (VANETs)
- Fuzzy Logic and Control Systems
- Algorithms and Data Compression
- Privacy, Security, and Data Protection
- Video Coding and Compression Technologies
- Anomaly Detection Techniques and Applications
- Swearing, Euphemism, Multilingualism
- Blockchain Technology Applications and Security
- Advanced Algorithms and Applications
- Simulation-Based Education in Healthcare
- Computational Fluid Dynamics and Aerodynamics
- Microbial Metabolites in Food Biotechnology
- Physical Unclonable Functions (PUFs) and Hardware Security
- Human Mobility and Location-Based Analysis
- Complex Network Analysis Techniques
- Advanced Control Systems Design
- Food composition and properties
- Education and Learning Interventions
- Multimodal Machine Learning Applications
- Advanced Sensor and Control Systems
Henan University
2024
Johns Hopkins University
2024
Chinese University of Hong Kong, Shenzhen
2023
Guangzhou University
2023
Zhengzhou University
2021-2022
University of Macau
2022
Soochow University
2022
Beijing Normal University
2022
China University of Mining and Technology
2019
Chengdu University of Information Technology
2018
Abstract The Metaverse has been the centre of attraction for educationists quite some time. This field got renewed interest with announcement social media giant Facebook as it rebranding and positioning Meta. While several studies conducted literature reviews to summarize findings related in general, no study best our knowledge focused on systematically summarizing finding education. To cover this gap, conducts a systematic review It then applies both content bibliometric analysis reveal...
Mobile crowdsensing (MCS) has heated up and become a new paradigm of data collection. In the process task allocation MCS, users are often required to provide their own location information with server conveniently dispatch some suitable tasks them. However, it is possible for malicious servers infer sensitive based on user's such as home address or trajectory so which will cause serious privacy issues. Recently, differential (DP) promising protection scheme. existing DP schemes in MCS do not...
With the substantial increase in number of smart cars, vehicular ad hoc network (VANET), where data can be shared between vehicles to enrich existing vehicle services and improve driving safety, is gaining more attention, thus creating a efficient intelligent transportation system. Moreover, in-depth research development 6G AI technology further strengthen interconnection various entities VANET realize edge intelligence, which fundamentally enhances efficiency sharing. However, reliable...
Chen Zhang, Yang Yang, Jiahao Liu, Jingang Wang, Yunsen Xian, Benyou Dawei Song. Proceedings of the 61st Annual Meeting Association for Computational Linguistics (Volume 1: Long Papers). 2023.
Quantization has emerged as a promising technique for improving the memory and computational efficiency of large language models (LLMs). Though trade-off between performance is well-known, there still much to be learned about relationship quantization LLM performance. To shed light on this relationship, we propose new perspective quantization, viewing it perturbations added weights activations LLMs. We call approach "the lens perturbation". Using lens, conduct experiments with various...
Quantization has emerged as a promising technique for improving the memory and computational efficiency of large language models (LLMs). Though trade-off between performance is well-known, there still much to be learned about relationship quantization LLM performance. To shed light on this relationship, we propose new perspective quantization, viewing it perturbations added weights activations LLMs. We call approach ``the lens perturbation". Using lens, conduct experiments with various...
Successive approximation register analog-to-digital converter (SAR ADC) is widely adopted in the Internet of Things (IoT) systems due to its simple structure and high energy efficiency. Unfortunately, SAR ADC dissipates various unique power features when it converts different input signals, leading severe vulnerability side-channel attack (PSA). The adversary can accurately derive signal by only measuring information from analog supply pin (AVDD), digital (DVDD), and/or reference (Ref) which...
With the increasing development of wireless communication technology and Vehicular Ad hoc Network (VANET), as well continuous popularization various sensors, Mobile Crowdsensing (MCS) paradigm has been widely concerned in field transportation. As a currently popular data sensing way, it mainly relies on devices to complete large‐scale complex tasks. However, since vehicles are highly mobile this scenario system is open, that is, any vehicle equipped with device can join system, credibility...
With the popularity of mobile smart devices, collecting data has become more convenient. Freeing from constraints professional equipment, Mobile Crowd Sensing (MCS) gained wide attention. As key component MCS system, task recommendation directly affects quantity and quality completion. However, most previous modules in system only consider situation where users complete tasks independently, without further consideration possibility that can seek assistance through social networks. In this...
Recently, there has been increasing interest in exploring the capabilities of advanced large language models (LLMs) field information extraction (IE), specifically focusing on tasks related to named entity recognition (NER) and relation (RE). Although researchers are use few-shot through in-context learning with LLMs, they tend focus only using correct or positive examples for demonstration, neglecting potential value incorporating incorrect negative into process. In this paper, we present...
Large language models (LLMs) have recently shown remarkable performance across a wide range of tasks. However, the substantial number parameters in LLMs contributes to significant latency during model inference. This is particularly evident when utilizing autoregressive decoding methods, which generate one token single forward process, thereby not fully capitalizing on parallel computing capabilities GPUs. In this paper, we propose novel approach, namely \textit{hidden transfer}, decodes...
The recent advancements in large language models (LLMs) have been extraordinary, yet the escalating inference costs associated with them present challenges real-world applications. To address these challenges, we propose a novel approach called Early-exiting Speculative Decoding (EESD) lossless acceleration. Specifically, EESD utilizes segment of LLM to generate draft tokens, incorporating structures after first N layers. enhance quality self-distillation method is integrated. This...
In this work, we demonstrate that physics-informed neural networks (PINNs) tend to propagate predicted shock wave information bidirectionally in time, which does not align with the actual evolution direction of solutions hyperbolic conservation laws. This mismatch results instability and hinders reduction loss governing equations, as well initial condition by deep network. order tackle problem, simplify complexity problem constructing equivalent linear transport equations region generation....
In this paper, an uncertain inverted pendulum system is controlled by IT2 fuzzy logic controller (IT2 FLC) which was optimized the particle swarm optimization (PSO) algorithm. The FLC, as extension of type-1 (T1 FLC), has better control performance in dealing with uncertainties than T1 FLC. However, FLC more complexity because parameters are so it difficult to select appropriate get performance. PSO algorithm used optimize several via two cost functions. According simulations, without...
Traditional PID control is widely used in industry due to its advantages of stability. The traditional controller very difficult obtain satisfied performance for complex nonlinear systems with uncertainties. Fuzzy approach has better anti-interference ability than method, and thus it can be deal many systems. In this paper, a comparative study the PID, type-1 type-2 fuzzy approaches are considered. By numerical simulations, that some specific situations, certain problem.
Recent studies have uncovered that language model distillation is less effective when facing a large capacity gap between the teacher and student, introduced assistant-based to bridge gap. As connection, scale performance of assistant vital importance bring knowledge from student. However, existing methods require maximally many trials before scheduling an optimal assistant. To this end, we propose minimal schedule (MiniDisc) for in minimally one trial. In particular, motivated by finding...
Versatile Video Coding (VVC) introduces more coding tools to improve compression efficiency compared its predecessor High Efficiency (HEVC). For inter-frame coding, Fractional Motion Estimation (FME) still has a high computational effort, which limits the real-time processing capability of video encoder. In this context, paper proposes an error-surface-based FME algorithm and corresponding hardware implementation. The creates error surface constructed by Rate-Distortion (R-D) cost integer...
Pretrained language models (LMs) have shown compelling performance on various downstream tasks, but unfortunately they require a tremendous amount of inference compute. Knowledge distillation finds path to compress LMs small ones with teacher-student paradigm. However, when the capacity gap between teacher and student is large, curse appears, invoking deficiency in distilling LMs. While few studies been carried out fill gap, not yet well tackled. In this paper, we aim at lifting via...
In today's digital economy, data sharing has become imperative for advancing economic growth and technology development, as multi-party collaboration requires secure exchange. During sharing, upholding security privacy is paramount. However, existing solutions mostly rely on cloud storage transfer, which have risks of single-point failure compromise security. To tackle these challenges, we propose a novel solution that leverages proxy re-encryption watermarking technologies, employs...