A5101756423- Data Mining Algorithms and Applications
- Rough Sets and Fuzzy Logic
- Bayesian Modeling and Causal Inference
- Machine Learning and Data Classification
- Face and Expression Recognition
- Data Management and Algorithms
- Data Quality and Management
- Imbalanced Data Classification Techniques
- Anomaly Detection Techniques and Applications
- Recommender Systems and Techniques
- Artificial Immune Systems Applications
- Semantic Web and Ontologies
- Text and Document Classification Technologies
- Image Retrieval and Classification Techniques
- Web Data Mining and Analysis
- AI-based Problem Solving and Planning
- Data Stream Mining Techniques
- Service-Oriented Architecture and Web Services
- Advanced Computational Techniques and Applications
- Algorithms and Data Compression
- Advanced Database Systems and Queries
- Advanced Image and Video Retrieval Techniques
- ECG Monitoring and Analysis
- Advanced Statistical Methods and Models
- Neural Networks and Applications
Changchun University
2024
Oklahoma State University
2002-2023
United Imaging Healthcare (China)
2023
Deakin University
2012-2022
Istituto Tecnico Industriale Alessandro Volta
2021
Second Affiliated Hospital of Zhengzhou University
2019-2021
Weatherford College
2021
Zhengzhou University
2020-2021
DePaul University
2001-2011
Microsoft (United States)
2005-2006
To engage visitors to a Web site at very early stage (i.e., before registration or authentication), personalization tools must rely primarily on clickstream data captured in server logs. The lack of explicit user ratings as well the sparse nature and large volume such setting poses serious challenges standard collaborative filtering techniques terms scalability performance. usage mining clustering that offline pattern discovery from transactions can be used improve filtering, however, this...
The appearance of generative adversarial networks (GAN) provides a new approach and framework for computer vision. Compared with traditional machine learning algorithms, GAN works via training concept is more powerful in both feature representation. also exhibits some problems, such as non-convergence, model collapse, uncontrollability due to high degree freedom. How improve the theory apply it computer-vision-related tasks have now attracted much research efforts. In this paper, recently...
Age Specific Human-Computer Interaction (ASHCI) has vast potential applications in daily life. However, automatic age estimation technique is still underdeveloped. One of the main reasons that aging effects on human faces present several unique characteristics which make a challenging task requires non-standard classification approaches. According to speciality facial effects, this paper proposes AGES (AGing pattErn Subspace) method for estimation. The basic idea model pattern, defined as...
Understanding goals and preferences behind a user's online activities can greatly help information providers, such as search engine E-Commerce web sites, to personalize contents thus improve user satisfaction. intention could also provide other business advantages providers. For example, providers decide whether display commercial content based on intent purchase. Previous work Web defines three major types of for queries: navigational, informational transactional or resource [1][7]. In this...
To engage visitors to a Web site at very early stage (i.e., before registration or authentication), personalization tools must rely primarily on clickstream data captured in server logs. The lack of explicit user ratings as well the sparse nature and large volume such setting poses serious challenges standard collaborative filtering techniques terms scalability performance. usage mining clustering that offline pattern discovery from transactions can be used improve filtering, however, this...
The KDD-Cup 2005 Competition was held in conjunction with the Eleventh ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. task of competition to classify 800,000 internet user search queries into 67 predefined categories. This is easy understand, but lack straightforward training set, subjective intents queries, poor information short high noise level make very challenge.In this paper, we summarize task, evaluation method, results competition. Here only highlight...
Arrhythmia is a disease that threatens human life. Therefore, timely diagnosis of arrhythmia great significance in preventing heart and sudden cardiac death. The BiLSTM-Attention neural network model with heartbeat activity's global sequence features can effectively improve the accuracy classification. Firstly, noise removed by continuous wavelet transform method. Secondly, peak R wave detected tagged database, then P-QRS-T morphology RR interval are extracted. This feature set features,...
Compared with conventional two-class learning schemes, one-class classification simply uses a single class in the classifier training phase. Applying to learn from unbalanced data set is regarded as recognition based and has shown have potential of achieving better performance. Similar learning, parameter selection significant issue, especially when sensitive parameters. For scheme kernel function, such Support Vector Machine Data Description, besides parameters involved kernel, there...
Nearest neighbor (NN) classification relies on the assumption that class conditional probabilities are locally constant. This becomes false in high dimensions with finite samples due to curse of dimensionality. The NN rule introduces severe bias under these conditions. We propose a adaptive neighborhood morphing method try minimize bias. use local support vector machine learning estimate an effective metric for producing neighborhoods elongated along less discriminant feature and constricted...
Test bar charts hidden in a 5.5-cm-thick 2.5% Intralipid solution were imaged as function of phantom depth and size with steady-state Fourier picosecond Kerr–Fourier imaging systems. With time space gating, series 250-μm bars placed thick highly scattering medium resolved at signal level ~10−10 the illumination intensity.
Nearest neighbor classification assumes locally constant class conditional probabilities. This assumption becomes invalid in high dimensions due to the curse-of-dimensionality. Severe bias can be introduced under these conditions when using nearest rule. We propose an adaptive method try minimize bias. use quasiconformal transformed kernels compute neighborhoods over which probabilities tend more homogeneous. As a result, better performance expected. The efficacy of our is validated and...
Myocardial infarction (MI) is an acute disease. Early detection and early treatment are of great significance for improving the health people. In order to reduce misdiagnosis rate MI diseases, this paper proposes a multi-lead bidirectional gated recurrent unit neural network (ML-BiGRU) learning algorithm based on current research status in field intelligent medical diagnosis, combined with timing correlation characteristics electrocardiogram (ECG) signals. At first, original ECG signal...
Most methods for classifying data streams operate under the hypothesis that distribution of classes is balanced. Unfortunately, phenomenon class imbalance widely exists in many real-world applications. In addition, underlying concept stream may change a certain way over time, and attacks increase difficulty mining. Motivated by this challenge, Two-Stage Cost-Sensitive (TSCS) classification proposed addressing issue non-stationary streams. We propose novel two-stage cost-sensitive framework...
This study investigates the potential of methyl-PEG2000-DSPE-PVP-LDC as a drug delivery nanocarrier and its impact on human immortalized keratinocytes, focusing cytotoxicity, migration inhibition, drug-loading efficiency. Synthesized nanoparticles were characterized using scanning electron microscopy, transmission zeta analysis, Fourier-transform infrared spectroscopy (FTIR). The cytotoxicity in keratinocyte HaCaT cells inhibition cell analyzed scratch assay. Furthermore, efficiency was...
Abstract Background TOAST subtype classification is important for diagnosis and research of ischemic stroke. Limited by experience neurologist time-consuming manual adjudication, it a big challenge to finish effectively. We propose novel active deep learning architecture classify TOAST. Methods To simulate the process neurologists, we drop valueless features XGB algorithm rank remaining ones. Utilizing framework, causal CNN, in which combines with mixed selection criterion optimize...
Nearest neighbor classification relies on the assumption that class conditional probabilities are locally constant. This becomes false in high dimensions with finite samples due to curse of dimensionality. The nearest rule introduces severe bias under these conditions. We propose a adaptive neighborhood morphing method try minimize bias. use local support vector machine teaming estimate an effective metric for producing neighborhoods elongated along less discriminant feature and constricted...