Yuxin Wang

ORCID: 0000-0003-4562-4055
Publications
Citations
Views
---
Saved
---
About
Contact & Profiles
Research Areas
  • Web Data Mining and Analysis
  • Privacy-Preserving Technologies in Data
  • Cryptography and Data Security
  • Advanced Computational Techniques and Applications
  • Service-Oriented Architecture and Web Services
  • Stochastic Gradient Optimization Techniques
  • Big Data Technologies and Applications
  • Handwritten Text Recognition Techniques
  • Cloud Computing and Resource Management
  • Spam and Phishing Detection
  • Interconnection Networks and Systems
  • Parallel Computing and Optimization Techniques
  • Text and Document Classification Technologies
  • Distributed and Parallel Computing Systems
  • Advanced Neural Network Applications
  • Advanced Memory and Neural Computing
  • Network Security and Intrusion Detection
  • Embedded Systems Design Techniques
  • IoT and Edge/Fog Computing
  • Digital Media Forensic Detection
  • Multimodal Machine Learning Applications
  • Semantic Web and Ontologies
  • Natural Language Processing Techniques
  • Radical Photochemical Reactions
  • Digital Marketing and Social Media

Beihang University
2024-2025

Harbin Institute of Technology
2014-2025

Korea Aerospace University
2025

Qilu University of Technology
2024

Shandong Academy of Sciences
2024

Hangzhou Normal University
2023-2024

National University of Defense Technology
2023

Chongqing University
2023

University of Science and Technology of China
2021-2023

University of Electronic Science and Technology of China
2023

Synthesis of bicyclo[1.1.1]pentane (BCP) heteroaryls continues to be a part the most important tasks in organic synthesis because they are significant class bioisosteres with universal applications drug discovery. However, substrate scope current multicomponent reactions is limited tertiary alkyl radicals and prefunctionalized (het)arenes due their intrinsic mechanisms, resulting decrease application value. Herein, we report straightforward alternative for (halo)alkyl BCP-heteroaryls from...

10.1021/acscatal.3c05003 article EN ACS Catalysis 2023-12-28

Distributed synchronous stochastic gradient descent (S-SGD) with data parallelism has been widely used in training large-scale deep neural networks (DNNs), but it typically requires very high communication bandwidth between computational workers (e.g., GPUs) to exchange gradients iteratively. Recently, Top-k sparsification techniques have proposed reduce the volume of be exchanged among and thus alleviate network pressure. can zero-out a significant portion without impacting model...

10.1109/icdcs.2019.00220 article EN 2019-07-01

In recent years, natural language processing (NLP) has got great development with deep learning techniques. the sub-field of machine translation, a new approach named Neural Machine Translation (NMT) emerged and massive attention from both academia industry. However, significant number researches proposed in past several there is little work investigating process this technology trend. This literature survey traces back origin principal timeline NMT, investigates important branches,...

10.48550/arxiv.2002.07526 preprint EN other-oa arXiv (Cornell University) 2020-01-01

This study describes a pioneering visible-light-induced phosphine-catalyzed halogen-atom transfer (XAT) strategy that heralds new era in the difunctionalization of [1.1.1]propellane.

10.1039/d4qo01351d article EN Organic Chemistry Frontiers 2024-01-01

Memory partitioning is widely adopted to efficiently increase the memory bandwidth by using multiple banks and reducing data access conflict. Previous methods for mainly focused on one-dimensional arrays. As a consequence, designers must flatten multidimensional array fit those methodologies. In this work we propose an automatic scheme arrays based linear transformation provide high throughput of on-chip memories loop pipelining in high-level synthesis. An optimal solution Ehrhart points...

10.1145/2463209.2488748 article EN 2013-05-28

The widespread acceptance of differential privacy has led to the publication many sophisticated algorithms for protecting privacy. However, due subtle nature this definition, such have bugs that make them violate their claimed In paper, we consider problem producing counterexamples incorrect algorithms. are designed be short and human-understandable so counterexample generator can used in development process -- a developer could quickly explore variations an algorithm investigate where they...

10.1145/3243734.3243818 preprint EN Proceedings of the 2022 ACM SIGSAC Conference on Computer and Communications Security 2018-10-15

This research examines how interpersonal rapport influences sustainable customer relationships during service failures, with a focus on empathy as mediating factor and emotional intelligence authenticity moderating variables. In the context of management, analysis data from 672 respondents through structural equation modeling reveals that significantly affects both cognitive dimensions empathy, subsequently enhancing satisfaction long-term loyalty outcomes. study demonstrates moderates...

10.3390/su17062396 article EN Sustainability 2025-03-09

Scene text editing (STE) aims to replace with the desired one while preserving background and styles of original text. However, due complicated textures various styles, existing methods fall short in generating clear legible edited images. In this study, we attribute poor performance two problems: 1) Implicit decoupling structure. Previous whole image have learn different translation rules regions simultaneously. 2) Domain gap. Due lack real scene images, network can only be well trained on...

10.1609/aaai.v37i2.25305 article EN Proceedings of the AAAI Conference on Artificial Intelligence 2023-06-26

The spread of DeepFake videos causes a serious threat to information security, calling for effective detection methods distinguish them. However, the performance recent frame-based become limited due their ignorance inter-frame inconsistency fake videos. In this paper, we propose novel Dynamic Inconsistency-aware Network handle inconsistent problem, which uses Cross-Reference module (CRM) capture both global and local inconsistencies. CRM contains two parallel branches. first branch takes...

10.24963/ijcai.2021/102 article EN 2021-08-01

Recent work on formal verification of differential privacy shows a trend toward usability and expressiveness -- generating correctness proof sophisticated algorithm while minimizing the annotation burden programmers. Sometimes, combining those two requires substantial changes to program logics: one recent paper is able verify Report Noisy Max automatically, but it involves complex system using customized logics verifiers. In this paper, we propose new technique, called shadow execution,...

10.1145/3314221.3314619 preprint EN 2019-06-07

Large language models (LLMs), especially Generative Pretrained Transformer (GPT) models, have significantly advanced in the industry recent years. However, these models' broader development faces considerable challenges due to high operational and deployment costs. This has led active research improving hardware efficiency of LLMs. Yet, characteristics real-world LLM workloads are often overlooked current optimizations serving systems. In this work, we find that absence reliable workload...

10.48550/arxiv.2401.17644 preprint EN arXiv (Cornell University) 2024-01-31

Behavioral synthesis tools have made significant progress in compiling high-level programs into register-transfer level (RTL) specifications. But manually rewriting code is still necessary order to obtain better quality of results memory system optimization. In recent years different automated optimization techniques been proposed and implemented, such as data reuse partitioning, but the problem integrating these an applicable flow a performance has become challenge. this paper we integrate...

10.1109/aspdac.2012.6164955 article EN 2012-01-01

We propose CheckDP, the first automated and integrated approach for proving or disproving claims that a mechanism is differentially private. CheckDP can find counterexamples mechanisms with subtle bugs which prior counterexample generators have failed. Furthermore, it was able to \emph{automatically} generate proofs correct no formal verification reported before. built on static program analysis, allowing be more efficient precise in catching infrequent events than existing (which run...

10.1145/3372297.3417282 article EN Proceedings of the 2022 ACM SIGSAC Conference on Computer and Communications Security 2020-10-30

In the study of urban crime, different region partitioning schemes often lead to analytical results, affecting accuracy and reliability research. This paper proposes a new method for crime data with multi-density distribution, namely Adaptive Region Partitioning (ARP) method, explores impact built environment features under methods on analysis. Compared traditional research unit ARP has better intra-regional homogeneity inter-regional heterogeneity characteristics, more accurately maps...

10.1109/access.2025.3525660 article EN cc-by IEEE Access 2025-01-01

10.1109/twc.2025.3574255 article EN IEEE Transactions on Wireless Communications 2025-01-01

Achieving optimal throughput by extracting parallelism in behavioral synthesis often exaggerates memory bottleneck issues. Data partitioning is an important technique for increasing bandwidth scheduling multiple simultaneous accesses to different banks. In this paper we present a vertical and algorithm that can generate valid partition scheme arbitrary affine inputs. It does arranging non-conflicting across the border of loop iterations. A mixed also proposed combine advantages other...

10.1145/2429384.2429484 article EN 2012-11-05

The traditional data analysis are based on the cause and effect relationship, formed a sample microscopic analysis, qualitative quantitative thinking mode of trend extrapolation analysis. Big has fundamental impact correlation, global macro technical correlation new Namely, from causal to knowledge discovery, model fitting mining, logical reasoning association rules. Data in era big have taken great changes, objects, processing, analytical methods tools, thinking.

10.1109/icbda.2016.7509810 article EN 2016-03-01

Various types of services, such as web APIs, IoT O2O and many others, have flooded on the Internet. Interconnections among these services resulted in a new phenomenon called "Internet Services" (IoS). By IoS, people don't need to request multiple by themselves fulfill their daily requirements, but it is an IoS platform that responsible for constructing integrated solutions them. Since user requirements (URs) are usually coarse-grained transboundary, platforms integrate from domains...

10.1109/scc49832.2020.00065 article EN 2020-11-01

As the big data Era, amount of in education shows explosive growth. Big analysis system architecture intends to provide a reference for large educational analysis. Through literature and network investigation, This paper reviewed existing mature system. From two aspects generality difference compare, this proposed research framework from several aspects, including common thinking, open source thought, areas, technology framework, core special functions. compares 3 mainstream system,...

10.1109/icbda.2017.8078706 article EN 2017-03-01

Noisy Max and Sparse Vector are selection algorithms for differential privacy serve as building blocks more complex algorithms. In this paper we show that both can release additional information free (i.e., at no cost). is used to return the approximate maximizer among a set of queries. We it also noisy gap between runner-up. This improve accuracy certain subsequent counting queries by up 50%. approximately larger than fixed threshold. adaptively control its budget (use less likely be much...

10.14778/3368289.3368295 article EN Proceedings of the VLDB Endowment 2019-11-01

Skywave over-the-horizon (OTH) radar systems have important long-range strategic warning values. They exploit skywave propagation reflection of high frequency signals from the ionosphere, which provides ultra-long-range surveillance capabilities to detect and track maneuvering targets. Current OTH are capable localizing targets in range azimuth but unable achieve reliable instantaneous altitude estimation. Most existing height measurement methods taken advantage micromultipath effect been...

10.1155/2014/424191 article EN cc-by Journal of Applied Mathematics 2014-01-01
Coming Soon ...