A5105759393- Neural Networks and Applications
- Advanced Image and Video Retrieval Techniques
- Face and Expression Recognition
- Image Retrieval and Classification Techniques
- Turbomachinery Performance and Optimization
- Aerodynamics and Acoustics in Jet Flows
- Fluid Dynamics and Turbulent Flows
- Fuzzy Logic and Control Systems
- Video Surveillance and Tracking Methods
- Rough Sets and Fuzzy Logic
- Combustion and flame dynamics
- Anomaly Detection Techniques and Applications
- Machine Learning and ELM
- Heat Transfer Mechanisms
- Imbalanced Data Classification Techniques
- Network Security and Intrusion Detection
- Computational Fluid Dynamics and Aerodynamics
- Multi-Criteria Decision Making
- Machine Learning and Data Classification
- Energy Load and Power Forecasting
- Data Mining Algorithms and Applications
- Text and Document Classification Technologies
- Aerodynamics and Fluid Dynamics Research
- Advanced Neural Network Applications
- Fault Detection and Control Systems
South China University of Technology
2016-2025
Virginia Tech
2016-2025
Goodwin College
2024-2025
University of South Australia
2024
Key Laboratory of Guangdong Province
2019-2023
Universidad Tecnológica de Panamá
2023
Techsburg (United States)
2003-2020
Engineering Systems (United States)
2019
Machine Science
2019
Education University of Hong Kong
2019
Rough sets and fuzzy have been proved to be powerful mathematical tools deal with uncertainty, it soon raises a natural question of whether is possible connect rough sets. The existing generalizations are all based on special relations (fuzzy similarity relations, T-similarity relations), advantageous generalize the by means arbitrary present general framework for study using both constructive axiomatic approaches. In this paper, from viewpoint approach, we first propose some definitions...
Fuzzy rough sets are the generalization of traditional to deal with both fuzziness and vagueness in data. The existing researches on fuzzy mainly concentrated construction approximation operators. Less effort has been put attributes reduction databases sets. This paper focuses After analyzing previous works sets, we introduce formal concepts completely study structure reduction. An algorithm using discernibility matrix compute all reductions is developed. Based these lines thought, set up a...
Undersampling is a widely adopted method to deal with imbalance pattern classification problems. Current methods mainly depend on either random resampling the majority class or at decision boundary. Random-based undersampling fails take into consideration informative samples in data while boundary sensitive overlapping. Both techniques ignore distribution information of training dataset. In this paper, we propose diversified sensitivity-based method. Samples are clustered capture and enhance...
Aim to currently content-based image retrieval method having high computational complexity and low accuracy problem, this paper proposes a based on color texture features. As its features, moments of the Hue, Saturation Value (HSV) component images in HSV space are used. Gabor descriptors adopted. Users assign weights each feature respectively calculate similarity with combined features according normalized Euclidean distance. Experiment results show that proposed has higher than...
The generalization error bounds found by current models using the number of effective parameters a classifier and training samples are usually very loose. These intended for entire input space. However, support vector machine (SVM), radial basis function neural network (RBFNN), multilayer perceptron (MLPNN) local learning machines solving problems treat unseen near to be more important. In this paper, we propose localized model which from above within neighborhood stochastic sensitivity...
This paper discusses the effects of multivariate correlation analysis on DDoS detection and proposes an example, a covariance model for detecting SYN flooding attacks. The simulation results show that this method is highly accurate in malicious network traffic attacks different intensities. can effectively differentiate between normal attack traffic. Indeed, detect even very subtle only slightly from behaviors. linear complexity makes its real time practical. to some extent verifies...
Fuzzy decision tree induction is an important way of learning from examples with fuzzy representation. Since the construction optimal NP-hard, research on heuristic algorithms necessary. In this paper, three for generating trees are analyzed and compared. One them proposed by authors. The comparisons twofold. analytic comparison based expanded attribute selection reasoning mechanism; other experimental size generated accuracy. purpose study to explore comparative strengths weaknesses...
In this paper, we focus on data-driven approaches to human activity recognition (HAR). Data-driven rely good quality data during training, however, a shortage of high quality, large-scale, and accurately annotated HAR datasets exists for recognizing activities daily living (ADLs) within smart environments. The contributions paper involve improving the an openly available dataset purpose proposing new ensemble neural networks as classifier. Specifically, propose homogeneous network approach...
The training of a multilayer perceptron neural network (MLPNN) concerns the selection its architecture and connection weights via minimization both error penalty term. Different terms have been proposed to control smoothness MLPNN for better generalization capability. However, controlling using, instance, norm or Vapnik-Chervonenkis dimension cannot distinguish individual MLPNNs with same number free parameters norm. In this paper, enhance capabilities, we propose stochastic sensitivity...
Auxetic plied yarns are specially constructed with two types of single different sizes and moduli. This paper investigates how to use these produce woven fabrics auxetic effects. Four-ply were first incorporated into a series design parameters study their behavior percent open area during extension. Effects yarn arrangement, component properties, weft type, weave structure then evaluated. Additional double helical (DHY) 6-ply also made for comparison. The results show that the alternative...
Current methods for nonintrusive load monitoring (NILM) problems assume that the number of appliances in target location is known, however, this may not be realistic. In real-world situations, initial setup site can known but new added by users after a period time, especially household or nonrestrictive scenarios. sense, current without detecting accurately monitor loads different and paper, novel appliance detection method proposed NILM with imbalance classification switching ON OFF. The...
Due to the fast, dynamic, and continuous arrival of data streams in green Internet Things (IoT) environment, probability distribution changes over time. In real IoT scenarios such as unmanned aerial vehicle (UAV) detection smart light switch control, have reduced trained model's accuracy for problems classification, making it challenging detect UAV intruders predict whether energy-saving lamps buildings are on or off. this paper, an incremental ensemble classification method is proposed...
<h3>Importance</h3> Retinopathy of prematurity (ROP) is the leading cause childhood blindness worldwide. Prediction ROP before onset holds great promise for reducing risk blindness. <h3>Objective</h3> To develop and validate a deep learning (DL) system to predict occurrence severity 45 weeks' postmenstrual age. <h3>Design, Setting, Participants</h3> This retrospective prognostic study included 7033 retinal photographs 725 infants in training set 763 90 external validation set, along with 46...
The applications of fuzzy production rules (FPR) are rather limited if the relative degree importance each proposition in antecedent contributing to consequent (i.e., weight) is ignored or assumed be equal. Unfortunately, this case for many existing FPR and most expert system development shells environments offer no such functionality users incorporate different weights FPR. This paper proposes assign a weight parameter new rule evaluation method (FPREM) which generalizes traditional by...
An important issue in the design and implementation of a neural network is sensitivity its output to input weight perturbations. In this paper, we discuss most popular general feedforward networks-multilayer perceptron (MLP). The defined as mathematical expectation errors MLP due perturbations with respect all values given continuous interval. for single neuron discussed first an analytical expression that function absolute approximately derived. Then algorithm compute entire MLP. As...
In most models of mining fuzzy association rules, the items are considered to have equal importance. Due diverse human interest and preference for items, such do not work well in many situations. To improve models, we propose a method mine rules with weighted items. One major problems data research is development good measures discovered rules. The support confidence defined. Kohonen self-organized mapping used fuzzify numerical attributes into linguistic terms. A new rule algorithm, which...
If the given fact for an antecedent in a fuzzy production rule (FPR) does not match exactly with of rule, consequent can still be drawn by technique such as reasoning. Many existing reasoning methods are based on Zadeh's Compositional Rule Inference (CRI) which requires setting up relation between and part. There some other do use CRI. Among them, similarity-based methods, make degree similarity to draw conclusion, well known. In this paper, six compared analyzed. Two them newly proposed...