A5100602151- Recommender Systems and Techniques
- Advanced Fiber Optic Sensors
- Photonic and Optical Devices
- Face and Expression Recognition
- Machine Learning and Data Classification
- Data Stream Mining Techniques
- Advanced Graph Neural Networks
- Text and Document Classification Technologies
- Image Retrieval and Classification Techniques
- Photonic Crystal and Fiber Optics
- Aquaculture disease management and microbiota
- Rough Sets and Fuzzy Logic
- Data Mining Algorithms and Applications
- Video Surveillance and Tracking Methods
- Semiconductor Lasers and Optical Devices
- Caching and Content Delivery
- Advanced Bandit Algorithms Research
- Anomaly Detection Techniques and Applications
- Cold Atom Physics and Bose-Einstein Condensates
- Topic Modeling
- Land Use and Ecosystem Services
- Advanced Malware Detection Techniques
- Network Security and Intrusion Detection
- Time Series Analysis and Forecasting
- Speech and Audio Processing
Southwest University
2010-2025
Capital Medical University
2008-2025
East China Normal University
2018-2025
Xinjiang Medical University
2025
Guangxi University
2025
Viva Biotech (China)
2025
Chongqing University of Posts and Telecommunications
2022-2024
Ministry of Agriculture and Rural Affairs
2024
Second Military Medical University
2024
Chinese Academy of Sciences
2014-2024
Recommender systems (RSs) commonly adopt a user-item rating matrix to describe users' preferences on items. With users and items exploding, such is usually high-dimensional sparse (HiDS). Recently, the idea of deep learning has been applied RSs. However, current deep-structured RSs suffer from high computational complexity. Enlightened by forest, this paper proposes latent factor model (DLFM) for building RS an HiDS efficiently. Its main construct sequentially connecting multiple (LF) models...
How to accurately predict unknown quality-of-service (QoS) data based on observed ones is a hot yet thorny issue in Web service-related applications. Recently, latent factor (LF) model has shown its efficiency addressing this owing high accuracy and scalability. An LF can be improved by identifying user service neighborhoods geographical information. However, such information difficult acquire most applications with the considerations of security, identity privacy, commercial interests real...
In this paper, we propose an open source speech recognition toolkit called WeNet, in which a new two-pass approach named U2 is implemented to unify streaming and non-streaming endto-end (E2E) single model.The main motivation of WeNet close the gap between research deployment E2E models.WeNet provides efficient way ship automatic (ASR) applications real-world scenarios, difference advantage other toolkits.We develop hybird connectionist temporal classification (CTC)/attention architecture...
Online streaming feature selection (OSFS) has attracted extensive attention during the past decades. Current approaches commonly assume that space of fixed data instances dynamically increases without any missing data. However, this assumption does not always hold in many real applications. Motivated by observation, study aims to implement online from sparse features, i.e., features flow one with as instance count remains fixed. To do so, proposes a latent-factor-analysis-based...
In this paper, we present WenetSpeech, a multi-domain Mandarin corpus consisting of 10000+ hours high-quality labeled speech, 2400+ weakly and about 10000 unlabeled with 22400+ in total. We collect the data from YouTube Podcast, which covers variety speaking styles, scenarios, domains, topics noisy conditions. An optical character recognition (OCR) method is introduced to generate audio/text segmentation candidates for on corresponding video subtitles, while ASR transcription system used...
Territorial Space (TS) is characterized by its multifunctionality. The identification and management of Spatial Functions (TSFs) across multi-scale crucial for achieving the SDGs. However, previous studies have primarily concentrated on variations in TSFs within administrative or grid units at a single scale, with investigations remaining challenge. This study focuses typical karst region Guangxi province China develops Multi-Scale Fusion model (MSF) assessing employs coupling coordination...
We report a highly sensitive refractive index (RI) sensor based on three cascaded single-mode fiber tapers, in which weak taper is sandwiched between the two tapers to improve sensitivity of sensor. Experimental results show that device 0.286 nm for 0.01 RI change, about four times higher than normal two-cascaded-taper-based Mach-Zehnder interferometer. In addition, could be enhanced by tapering longer and thinner middle taper. Such kinds low-cost fiber-optic sensors would find applications...
In-line fiber optic interferometers have attracted intensive attention for their potential sensing applications in refractive index, temperature, pressure and strain measurement, etc. Typical in-line fiber-optic are of two types: Fabry-Perot core-cladding-mode interferometers. It's known that the based on single-mode fibers can exhibit compact structures, easy fabrication low cost. In this paper, we review kinds typical formed fabricated with different post-processing techniques. Also, some...
Non-negativity is vital for a latent factor (LF)-based model to preserve the important feature of high-dimensional and sparse (HiDS) matrix in recommender systems, i.e., none its entries negative. Current non-negative models rely on constraints-combined training schemes. However, they lack flexibility, scalability, or compatibility with general This work aims perform unconstrained analysis (UNLFA) HiDS matrices. To do so, we innovatively transfer non-negativity constraints from decision...
As an invasive alien species, Alternanthera philoxeroides has spread rapidly in China since its invasion, which caused serious harm to the ecological environment and substantial economic losses. To better manage this plant, we analysed spatial temporal distribution patterns predicted suitable areas for species using MaxEnt model ArcGIS. The results showed that area A. current climatic conditions was 91.8–122.7°E,18.2–39.8°N, mainly located tropical, subtropical southeastern warm temperate...
A peanut-shape fiber structure that can realize the coupling and recoupling between core mode cladding modes is proposed in this paper. Based on structure, a simple low-cost Mach-Zehnder interferometer (MZI) formed by cascading two structures single-mode demonstrated. The theory experimental results show first couple light energy of into second recouple mode. high-quality interference spectrum with fringe visibility about 13 dB observed. Moreover, it has very good mechanical strength...
The tunable refractive index of the magnetic fluid (MF) is a unique optical property, which has attracted lot research interest in recent years. In this paper, method based on Fresnel reflection at fiber end face presented. Experimental measurements are carried out to investigate field (intensity and direction) temperature-dependent MF. For given concentration, with increase intensity, nMF increases gradually when H//Light, while decreases H⊥Light. effect temperature relatively insignificant...
Neighborhood regularization is highly important for a latent factor (LF)-based Quality-of-Service (QoS)-predictor since similar users usually experience QoS when invoking services. Current neighborhood-regularized LF models rely prior information on neighborhood obtained from common raw data or geographical information. The former suffers low prediction accuracy due to the difficulty of constructing based incomplete data, while latter requires additional that difficult collect considering...
A compact optical fiber magnetic field sensor based on the principle of Sagnac interferometer is proposed. Different from conventional ones, a ferrofluid-filled high-birefringence photonic crystal (HB-PCF) inserted into as sensing element. The refractive index ferrofluid filled in cladding air holes HB-PCF will change with respect to applied field, and subsequently, birefringence change, which affect shifts output interference spectrum Sagnac. Experiments are carried out verify simulation...
Self-labeled technique, a paradigm of semisupervised classification (SSC), is highly effective in alleviating the shortage labeled data tasks via an iterative self-labeling process. Although existing self-labeled SSC models show great prospect industrial applications, they suffer from performance degeneration caused by false-positive label-predictions unlabeled during For addressing this issue, paper proposes novel framework, which compatible with most models. The main idea framework to...
A recommender system (RS) is highly efficient in filtering people's desired information from high-dimensional and sparse (HiDS) data. To date, a latent factor (LF)-based approach becomes popular when implementing RS. However, current LF models mostly adopt single distance-oriented Loss like an L2 norm-oriented one, which ignores target data's characteristics described by other metrics L1 one. investigate this issue, article proposes -and- -norm-oriented ( [Formula: see text]) model. It...
High-dimensional and sparse (HiDS) matrices commonly arise in various industrial applications, e.g., recommender systems (RSs), social networks, wireless sensor networks. Since they contain rich information, how to accurately represent them is of great significance. A latent factor (LF) model one the most popular successful ways address this issue. Current LF models mostly adopt L <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub>...
Performing highly accurate representation learning on a high-dimensional and sparse (HiDS) matrix is of great significance in big data-related application such as recommender system. A latent factor (LF) model one the most efficient approaches to HiDS representation. However, an LF model's ability relies heavily matrix's known data density, which extremely low due numerous missing entities. To address this issue, work proposes prediction-sampling-based multilayer-structured (PMLF) with...
Quality-of-Service (QoS), which describes the non-functional characteristics of Web service, is great significance in service selection. Since users cannot invoke all services to obtain corresponding QoS data, prediction becomes a hot yet thorny issue. To date, latent factor analysis (LFA)-based predictor one most successful and popular approaches address this However, current LFA-based predictors are mostly modeled on inner product space with an <italic...
Deep learning (DL)-based recommender system (RS), particularly for its advances in the recent five years, has been startling. It reshapes architectures of traditional RSs by lifting their limitations dealing with data sparsity and cold-start issues. Yet, performance DL-based RS, like many other intelligent systems, heavily relies on selecting hyperparameters. Unfortunately, most common selection approach is still Grid Search that requires numerous computational resources human efforts....
Wireless sensor network (WSN) is an emerging and promising developing area in the intelligent sensing field. Due to various factors like sudden sensors breakdown or saving energy by deliberately shutting down partial nodes, there are always massive missing entries collected data from WSNs. Low-rank matrix approximation (LRMA) a typical effective approach for pattern analysis recovery However, existing LRMA-based approaches ignore adverse effects of outliers inevitably mixed with data, which...
ABSTRACT Cerebral venous sinus thrombosis (CVST) is frequently observed in younger adults and features large thrombus volume. Due to the triangular-like cross-sectional shape diameter of superior sagittal sinus, all commercially available artery stent retrievers are not suitable for vessels. In this study, a dumbbell-like was designed fabricated by 3D braided technology using NiTi wires; it manually rotatable stretchable with controlled length/diameter ratios (2.6–14.0) reciprocating...