Xiaoyun Li

ORCID: 0000-0001-5730-2972
Publications
Citations
Views
---
Saved
---
About
Contact & Profiles
Research Areas
  • Software System Performance and Reliability
  • Energy Efficient Wireless Sensor Networks
  • Cloud Computing and Resource Management
  • Privacy-Preserving Technologies in Data
  • Mobile Ad Hoc Networks
  • Stochastic Gradient Optimization Techniques
  • Head and Neck Cancer Studies
  • Network Security and Intrusion Detection
  • Caching and Content Delivery
  • Recommender Systems and Techniques
  • IoT and Edge/Fog Computing
  • Cholangiocarcinoma and Gallbladder Cancer Studies
  • Fuel Cells and Related Materials
  • Bayesian Methods and Mixture Models
  • Wireless Networks and Protocols
  • Advanced Bandit Algorithms Research
  • Indoor and Outdoor Localization Technologies
  • Cryptography and Data Security
  • Energy Harvesting in Wireless Networks
  • Anomaly Detection Techniques and Applications
  • Blockchain Technology Applications and Security
  • Distributed Control Multi-Agent Systems
  • Sparse and Compressive Sensing Techniques
  • Underwater Vehicles and Communication Systems
  • Viral-associated cancers and disorders

Sun Yat-sen University
2006-2024

Sun Yat-sen University Cancer Center
2019-2024

Capital Institute of Pediatrics
2023

Hebei Medical University
2023

Fourth Hospital of Hebei Medical University
2023

Baidu (China)
2019-2022

China University of Labor Relations
2021

Beijing University of Posts and Telecommunications
2018-2021

Bellevue Hospital Center
2020-2021

Shandong University
2021

BackgroundPrevious mass screening studies have shown that IgA antibodies against Epstein–Barr Virus (EBV) can facilitate early detection of nasopharyngeal carcinoma (NPC), but the impact EBV-antibody for NPC-specific mortality remains unknown.Patients and methodsA prospective, cluster randomized, controlled trial NPC (PRO-NPC-001) was conducted in 3 selected towns Zhongshan City 13 Sihui southern China beginning 2008. Serum samples group were tested two previously anti-EBV antibodies....

10.1093/annonc/mdz231 article EN cc-by-nc-nd Annals of Oncology 2019-08-01

Anomaly detection is a critical task for maintaining the performance of cloud system. Using data-driven methods to address this issue mainstream in recent years. However, due lack labeled data training practice, it necessary enable an anomaly model trained on contaminated unsupervised way. Besides, with increasing complexity systems, effectively organizing collected from wide range components system and modeling spatiotemporal dependence among them become challenge. In article, we propose...

10.1109/tnnls.2020.3027736 article EN IEEE Transactions on Neural Networks and Learning Systems 2020-10-16

Log-based anomaly detection has been widely studied and achieves a satisfying performance on stable log data. But, the existing approaches still fall short meeting these challenges: 1) Log formats are changing continually in practice those software systems under active development maintenance. 2) Performance issues latent causes that may not be detected by trivial monitoring tools. We thus propose SwissLog, namely robust unified deep learning based model for detecting diverse faults....

10.1109/issre5003.2020.00018 article EN 2020-10-01

With the advantages of flexible scalability and fast delivery, microservice has become a popular software architecture in modern IT industry. However, explosion number service instances complex dependencies make troubleshooting extremely challenging environments. To help understand troubleshoot system, end-to-end tracing technology been widely applied to capture execution path each request. Nevertheless, data are not fully leveraged by cloud application providers when conducting latency...

10.1145/3442381.3449905 article EN 2021-04-19

10.1016/j.jpdc.2020.12.003 article EN Journal of Parallel and Distributed Computing 2021-01-01

Log data is a crucial resource for recording system events and states during execution. However, as systems grow in scale, log generation has become increasingly explosive, leading to an expensive overhead on storage, such several petabytes per day production. To address this issue, compression task reducing disk storage while allowing further analysis. Unfortunately, existing general-purpose log-specific methods have been limited their ability utilize characteristics. overcome these...

10.1145/3597503.3608129 article EN 2024-02-06

Modern distributed systems generate interleaved logs when running in parallel. Identifiers (ID) are always attached to them trace instances or entities logs. Therefore, log messages can be grouped by the same IDs help anomaly detection and localization. The existing approaches achieve this still fall short meeting these challenges: 1) Log is solely processed single components without mining dependencies. 2) formats continually changing modern software systems. 3) It challenging detect latent...

10.1109/tdsc.2022.3162857 article EN IEEE Transactions on Dependable and Secure Computing 2022-03-29

Root cause analysis (RCA) in large-scale microservice systems is a critical and challenging task. To understand localize root causes of unexpected faults, modern observability tools collect preserve multi-modal data, including metrics, traces, logs. Since system faults may manifest as anomalies different data sources, existing RCA approaches that rely on single-modal are constrained the granularity interpretability causes. In this study, we present Nezha, an interpretable fine-grained...

10.1145/3611643.3616249 article EN 2023-11-30

Cofilin 1 (CFL1) is a cytoskeletal protein and overexpression of the has been associated with aggressiveness in certain types malignancies. The aim present study was to investigate clinical implications CFL1 expression prostate cancer (PCa). Immunohistochemical analysis performed using formalin-fixed paraffin-embedded tissue sections obtained from 111 patients PCa 47 benign prostatic hyperplasia (BPH). In total, 78 (70.3%) out tissues were found express protein, while no detected BPH...

10.3892/ol.2015.3133 article EN Oncology Letters 2015-04-21

Deep neural network has been adopted as the standard model to predict ads click-through rate (CTR) for commercial online advertising systems. Deploying an industrial scale system requires overcome numerous challenges, e.g., hundreds or thousands of billions input features and also training samples, which under cost budget can cause fundamental issues on storage, communication, speed. In this work, we present Baidu's industrial-scale practices how apply machine learning techniques address...

10.1145/3448016.3457236 article EN Proceedings of the 2022 International Conference on Management of Data 2021-06-09

Abstract Background Currently, the diagnosis and treatment of nasopharyngeal carcinoma (NPC) patients with residual cervical lymphadenopathy following radical radiotherapy or without chemotherapy are challenging. We investigated prognosis NPC assessed diagnostic prognostic values Epstein‐Barr virus (EBV) DNA in these patients. Methods This study included 82 who were diagnosed suspected completion antitumor therapy. Their plasma EBV levels measured using quantitative polymerase chain reaction...

10.1186/s40880-019-0357-9 article EN cc-by Cancer Communications 2019-03-29

Abstract Background Capecitabine was previously used as a second‐line or salvage therapy for metastatic nasopharyngeal carcinoma (NPC) and has shown satisfactory curative effect maintenance in other cancers. This study aimed to explore the role of capecitabine de novo NPC patients with different plasma Epstein‐Barr virus (EBV) DNA levels before treatment. Methods We selected treated locoregional radiotherapy (LRRT) this retrospective study. The propensity score matching (PSM) applied balance...

10.1002/cac2.12004 article EN cc-by-nc-nd Cancer Communications 2020-01-01

An analytical approximation is developed for the probability of sensing coverage in a wireless sensor network with randomly deployed nodes each having an isotropic area. This approximate obtained by considering properties geometric graph, which edge exists between any two vertices representing overlapping areas. The principal result to proportion area that covered at least one node, given expected number per unit two-dimensional Poisson process. specified region being completely also...

10.1109/tit.2011.2169300 article EN IEEE Transactions on Information Theory 2012-01-01

In recent years, distributed optimization is proven to be an effective approach accelerate training of large scale machine learning models such as deep neural networks. With the increasing computation power GPUs, bottleneck speed in gradually shifting from communication. Meanwhile, hope on mobile devices, a new paradigm called "federated learning'' has become popular. The communication time federated especially important due low bandwidth devices. While various approaches improve efficiency...

10.1145/3412815.3416891 article EN 2020-10-15

Serverless Function-as-a-Service (FaaS) is a rapidly growing computing paradigm in the cloud era. To provide rapid service response and save network bandwidth, traditional cloud-based FaaS platforms have been extended to edge. However, launching functions heterogeneous continuum (HCC) that includes cloud, fog, edge brings new challenges: determining <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">where should be delivered how many resources...

10.1109/tsc.2023.3274769 article EN IEEE Transactions on Services Computing 2023-05-11

Abstract Background The coronavirus disease 2019 outbreak has hit Beijing since mid-Nov, 2022, with soaring growth of severe acute respiratory syndrome 2 (SARS-CoV-2) among children. Therefore, it is vital to determine the clinical manifestations epidemic SARS-CoV-2 strains in paediatric patients. Methods In this study, nucleic acid tests (NATs) for were performed outpatients symptoms tract infection during 18 Nov–6 Dec, 2022. Half positive randomly selected screen other pathogens, whereas...

10.1186/s12985-023-02177-x article EN cc-by Virology Journal 2023-09-08

This paper shows how a distributed algorithm, derived and justified through computational geometry, can detect recover holes in the coverage provided by wireless sensor networks. Computational geometry is used to define conditions for existence of based on these conditions, it generate new algorithms holes. The algorithm does not require coordinates or location information requires only minimal connectivity information. Most be detected with very low probability error simulation results...

10.1504/ijsnet.2009.026363 article EN International Journal of Sensor Networks 2009-01-01

Faults are the primary culprits of breaking high availability cloud systems, even leading to costly outages. As scale and complexity clouds increase, it becomes extraordinarily difficult understand, detect diagnose faults. During outages, engineers record detailed information whole life cycle faults (i.e., fault occurrence, detection, identification, mitigation) in form postmortems. In this paper, we conduct a quantitative qualitative study on 354 public post-mortems collected three popular...

10.1109/issre55969.2022.00022 article EN 2022-10-01

A distributed algorithm is introduced which detects and recovers holes in the coverage provided by wireless sensor networks. It does not require coordinates, requiring only minimal connectivity information (for example, whether any two nodes are within either sensing radius or twice radius.) The radio communications area assumed to be larger than sensed area. Two active called neighbors, said connected a link, if their distance lies between these values. Redundant likely exist inside an...

10.1109/glocom.2006.680 article EN Globecom 2006-11-01
Coming Soon ...