A5100872537- Advanced Graph Neural Networks
- Advanced Bandit Algorithms Research
- Recommender Systems and Techniques
- Industrial Vision Systems and Defect Detection
- Textile materials and evaluations
- Remote-Sensing Image Classification
- Remote Sensing and Land Use
- Infrared Target Detection Methodologies
- Evaluation Methods in Various Fields
- Sparse and Compressive Sensing Techniques
- Industrial Technology and Control Systems
- Advanced machining processes and optimization
- Service-Oriented Architecture and Web Services
- Color Science and Applications
- Advanced Measurement and Metrology Techniques
- Advanced Computational Techniques and Applications
- Blind Source Separation Techniques
- Iterative Learning Control Systems
- Spectroscopy and Quantum Chemical Studies
- Nanopore and Nanochannel Transport Studies
- Electrostatics and Colloid Interactions
- Face and Expression Recognition
- Evaluation and Optimization Models
Tsinghua University
2024
University Town of Shenzhen
2024
University of Chinese Academy of Sciences
2017-2023
Shanghai Institute of Applied Physics
2023
Tencent (China)
2020-2022
Chinese Academy of Sciences
2017
Institute of Electronics
2017
Tiangong University
2010
National Chung Cheng University
1996
Both explicit and implicit feedbacks can reflect user opinions on items, which are essential for learning preferences in recommendation. However, most current recommendation algorithms merely focus positive (e.g., click), ignoring other informative behaviors. In this paper, we aim to jointly consider explicit/implicit positive/negative learn unbiased Specifically, propose a novel Deep feedback network (DFN) modeling click, unclick dislike DFN has an internal interaction component that...
The linear operator has been widely used to detect targets of interest in multispectral/hyperspectral images, and is usually able achieve good performance when the target linearly separable from background. However, dealing with a complex scene, it hard find single projection direction, along which can be well distinguished all background objects. Therefore, we propose piecewise strategy (PLS) for detection, based on assumption that desired generally each object. PLS first divides whole into...
Personalized recommendation aims to provide appropriate items according user preferences mainly from their behaviors. Excessive homogeneous behaviors on similar will lead fatigue, which may decrease activeness and degrade experience. However, existing models seldom consider fatigue in recommender systems. In this work, we propose a novel multi-granularity modeling coarse fine. Specifically, focus the feed scenario, where underexplored global session coarse-grained taxonomy have large...
Experiments and theory have revealed versatile possible phases for adsorbed confined water on two-dimensional solid surfaces, which are closely related to the aspects of various phenomena in physics, chemistry, biology, tribology. In this review, we summarize our recent works showing that different with disordered ordered structures can greatly affect surface wetting behavior, dielectric properties, frictions. This includes phase structure induces an unexpected phenomenon, “ordered monolayer...
Based on the multiplicative update rule of nonnegative matrix factorization (NMF), we add constraint endmember to objective function iterative constrained endmembers (ICE), and propose a simple for ICE, named NMF-ICE. Our method avoids use quadratic programming in therefore can greatly improve computational efficiency. Experiments using both simulated real hyperspectral data show that NMF-ICE is effective generation.
Recommender systems filter out information that meets user interests. However, users may be tired of the recommendations are too similar to content they have been exposed in a short historical period, which is so-called fatigue. Despite significance for better experience, fatigue seldom explored by existing recommenders. In fact, there three main challenges addressed modeling fatigue, including what features support it, how it influences interests, and its explicit signals obtained. this...
In this paper, the method that measuring dataset of knitted yarns is clustered using improving fuzzy kernel c-Means (FKCM) clustering algorithm proposed. FKCM algorithm, data low dimension input space mapped to high feature space, FCM performed in then constraint optimization distance matrix and membership testing samples are computed by utilizing iterative result can be acquired according maximum principle. Subsequently, Kernel F cluster validity index designed for seeking fitness number...
This paper presents the support vector machine (SVM) for classification of quality grade knitted yarns. The SVM, Kernel Fisher Discriminant Analysis (KFDA), back promulgation neural network (BPNN), and radial basis function (RBFNN) are comparatively investigated in 94 classified yarns from different mills four-dimensional space, four methods employed on IRIS dataset, experimental results exhibit SVM method has best effect. FKCM can constitute a complete evaluation system, provide an...