A5100438478- Data Stream Mining Techniques
- Machine Learning and Data Classification
- Advanced Data Storage Technologies
- Network Security and Intrusion Detection
- Target Tracking and Data Fusion in Sensor Networks
- Imbalanced Data Classification Techniques
- Internet Traffic Analysis and Secure E-voting
- Advanced Malware Detection Techniques
- Anomaly Detection Techniques and Applications
- Advanced Memory and Neural Computing
- Parallel Computing and Optimization Techniques
- Domain Adaptation and Few-Shot Learning
- Multimodal Machine Learning Applications
- Natural Language Processing Techniques
- Topic Modeling
- Privacy-Preserving Technologies in Data
- Neural Networks and Applications
- Smart Grid Energy Management
- Infrared Target Detection Methodologies
- Advanced Bandit Algorithms Research
- Solar Radiation and Photovoltaics
- Text and Document Classification Technologies
- Telecommunications and Broadcasting Technologies
- Bioinformatics and Genomic Networks
- Forecasting Techniques and Applications
National University of Defense Technology
2012-2024
Shenyang Aerospace University
2023
Gannan Normal University
2023
China Southern Power Grid (China)
2022
University of California, Riverside
2022
Beijing University of Posts and Telecommunications
2022
Northwestern Polytechnical University
2019
Zhejiang Province Institute of Architectural Design and Research
2014
Zhejiang University
2014
Nanjing University
2012
In practical applications, data stream classification faces significant challenges, such as high cost of labeling instances and potential concept drifting. We present a new online active learning ensemble framework for drifting streams based on hybrid strategy that includes the following: 1) an classifier, which consists long-term stable classifier multiple dynamic classifiers (a multilevel sliding window model is used to create update effectively process both gradual drift type sudden...
A challenge to many real-world applications is multiclass imbalance with concept drift. In this paper, we propose a comprehensive active learning method for imbalanced streaming data drift (CALMID). First, design online framework that includes an ensemble classifier, detector, label sliding window, sample windows and initialization training sequence. Next, variable threshold uncertainty strategy based on asymmetric margin matrix designed comprehensively address the problem given class can...
The complex problems of multiclass imbalance, virtual or real concept drift, evolution, high-speed traffic streams and limited label cost budgets pose severe challenges in network classification tasks. In this paper, we propose a imbalanced drift framework based on online active learning (MicFoal), which includes configurable supervised learner for the initialization model, an method with hybrid request strategy, sliding window group, sample training weight formula adaptive adjustment...
Current dense text retrieval models face two typical challenges. First, they adopt a siamese dual-encoder architecture to encode queries and documents independently for fast indexing searching, while neglecting the finer-grained term-wise interactions. This results in sub-optimal recall performance. Second, their model training highly relies on negative sampling technique build up contrastive losses. To address these challenges, we present Adversarial Retriever-Ranker (AR2), which consists...
Compared to image-text pair data, interleaved corpora enable Vision-Language Models (VLMs) understand the world more naturally like humans. However, such existing datasets are crawled from webpage, facing challenges low knowledge density, loose relations, and poor logical coherence between images. On other hand, internet hosts vast instructional videos (e.g., online geometry courses) that widely used by humans learn foundational subjects, yet these valuable resources remain underexplored in...
In this paper, we propose VideoLLaMA3, a more advanced multimodal foundation model for image and video understanding. The core design philosophy of VideoLLaMA3 is vision-centric. meaning "vision-centric" two-fold: the vision-centric training paradigm framework design. key insight our that high-quality image-text data crucial both Instead preparing massive video-text datasets, focus on constructing large-scale datasets. has four stages: 1) alignment stage, which warms up vision encoder...
Practical applications often require learning algorithms capable of addressing data streams with concept drift and class imbalance. This paper proposes an online active paired ensemble for drifting The consists a long-term stable classifier dynamic to address both sudden gradual drift. To select the most representative instances learning, hybrid labeling strategy which includes uncertainty imbalance is proposed. applies margin-based criterion adjustment threshold. Based on categorical...
Machine learning in real-world scenarios is often challenged by concept drift and class imbalance. This paper proposes a Resample-based Ensemble Framework for Drifting Imbalanced Stream (RE-DI). The ensemble framework consists of long-term static classifier to handle gradual multiple dynamic classifiers sudden drift. weights the are adjusted from two aspects. First, time-decayed strategy decreases make focus more on new data stream. Second, novel reinforcement mechanism proposed increase...
Applications challenged by the joint problem of concept drift and class imbalance are attracting increasing research interest. This paper proposes a novel Reinforcement Online Active Learning Ensemble for Drifting Imbalanced data stream (ROALE-DI). The ensemble classifier has long-term stable dynamic group which applies reinforcement mechanism to increase weight classifiers, perform better on minority class, decreases opposite. When is imbalanced, classifiers will lack training samples...
To facilitate efficient context switches, GPUs usually employ a large-capacity register file to accommodate massive amount of information. However, the large introduces high power consumption, flowing leakage SRAM cells. Emerging non-volatile STT-RAM memory has recently been studied as potential replacement alleviate challenge when constructing files on GPUs. Unfortunately, due long write latency and energy consumption associated with operations in STT-RAM, simply replacing STTRAM for would...
Protecting confidential data against memory disclosure attacks is crucial to many critical applications, especially those involve cryptographic operations. However, it neither easy identify involved in a program nor implement fine-grained and yet efficient protection. Existing defensive techniques face shortcomings such as coarse-grained protection or exorbitant overhead. As result, real world crypto applications seldom applied this kind of practice.To make the practical, we design...
Many ontologies have been published on the Semantic Web, to be shared describe resources. Among them, large of real-world areas scalability problem in presenting semantic technologies such as ontology matching (OM). This either suffers from too long run time or has strong hypotheses running environment. To deal with this issue, we propose a three-stage MapReduce-based approach V-Doc+ for ontologies, based MapReduce framework and virtual document technique. Specifically, two processes are...
The biological activity predictions of ligands are an important research direction, which can improve the efficiency and success probability drug screening. However, traditional prediction method has disadvantages complex modeling low screening efficiency. Machine learning is considered direction to solve these problems in near future. This paper proposes a machine model with high predictive accuracy stable ability, namely, back propagation neural network cross-support vector regression...
Machine learning algorithms have been widely used in the field of client credit assessment. However, few focused on and solved problems concept drift class imbalance. Due to changes macroeconomic environment markets, relationship between characteristics assessment results may change over time, causing assessments. Moreover, data are naturally asymmetric imbalanced because screening clients. Aiming at solving joint research issue imbalance assessments, this paper, a novel sample-based online...
To address the high energy consumption issue of SRAM on GPUs, emerging Spin-Transfer Torque (STT-RAM) memory technology has been intensively studied to build GPU register files for better energy-efficiency, thanks its benefits low leakage power, density, and good scalability. However, STT-RAM suffers from read disturbance issue, which stems fact that voltage difference between current write becomes smaller as scales. The leads error rates operations, cannot be effectively protected by...
To address the high energy consumption issue of SRAM on GPUs, emerging Spin-Transfer Torque (STT-RAM) memory technology has been intensively studied to build GPU register files for better energy-efficiency, thanks its benefits low leakage power, density, and good scalability. However, STT-RAM suffers from a reliability issue, read disturbance, which stems fact that voltage difference between current write becomes smaller as scales. The disturbance leads error rates operations, cannot be...
For improving the detection of micro-calcifications (MCs), this paper proposes an automatic MC system making use multi-fractal spectrum in digitized mammograms. The approach is based on principle that
This paper proposes an air quality grade forecasting method based on ensemble learning. First, the training data sets are formed of and related meteorological crawled from website. After that, use learning algorithm Leveraging Bagging to learn dataset generate initial model. And model is used make prediction dataset. In total, experiments test both city scale station scale. Experimental results show that proposed has good effect ability real forecast
The complex problems of multiclass imbalance, virtual or real concept drift, evolution, high-speed traffic streams and limited label cost budgets pose severe challenges in network classification tasks. In this paper, we propose a m ulticlass i mbalanced c oncept drift f ramework based on o nline ctive l earning (MicFoal), which includes configurable supervised learner for the initialization model, an active learning method with hybrid request strategy, sliding window group, sample training...