A5060931310- Time Series Analysis and Forecasting
- Metaheuristic Optimization Algorithms Research
- Evolutionary Algorithms and Applications
- Anomaly Detection Techniques and Applications
- Advanced Multi-Objective Optimization Algorithms
- Conducting polymers and applications
- VLSI and FPGA Design Techniques
- Optical Network Technologies
- Data Management and Algorithms
- Complex Systems and Time Series Analysis
- Face and Expression Recognition
- Data Stream Mining Techniques
- Advanced Thermoelectric Materials and Devices
- Distributed and Parallel Computing Systems
- Topic Modeling
- Advanced Sensor and Energy Harvesting Materials
- Authorship Attribution and Profiling
- Parallel Computing and Optimization Techniques
- Advanced Fiber Laser Technologies
- Remote-Sensing Image Classification
- Polydiacetylene-based materials and applications
- Imbalanced Data Classification Techniques
- Microfluidic and Bio-sensing Technologies
- Nonlinear Photonic Systems
- Hate Speech and Cyberbullying Detection
Shenzhen University
2023-2025
Zhongnan University of Economics and Law
2023-2024
Zhengzhou University of Light Industry
2022-2024
Wuhan University
2014-2023
University of Edinburgh
2022
CRRC (China)
2020
Joint Laboratory for Extreme Conditions Matter Properties
2020
Southwest University of Science and Technology
2020
University of Houston
2011-2019
State Key Laboratory of Software Engineering
2010-2017
Differential evolution (DE) is an efficient and powerful stochastic optimization algorithm. Extensive studies in recent years have verified that different trial vector generation strategies associated control parameters offer distinct characteristics on problems. To take full advantages of them, ensemble methods based various adaptive been proposed during the last decade. Aiming to organically integrate merits some popular parameters, then utilize a multi-role DE (MRDE) this paper. In MRDE,...
Large language models (LLMs) are remarked by their substantial computational requirements. To mitigate the cost, researchers develop specialized CUDA kernels, which often fuse several tensor operations to maximize utilization of GPUs as much possible. However, those kernels may still leave performance on table assembly experts show that manual optimization GPU SASS schedules can lead better performance, and trial-and-error is largely employed manually find best schedules. In this work, we...
Recent advancements in Large Language Models (LLMs) have led to increasingly diverse requests, accompanied with varying resource (compute and memory) demands serve them. However, this turn degrades the cost-efficiency of LLM serving as common practices primarily rely on homogeneous GPU resources. In response problem, work conducts a thorough study about LLMs over heterogeneous resources cloud platforms. The rationale is that different types exhibit distinct compute memory characteristics,...
Filtering has been an enabling technology and found ever-increasing applications. There are two main classes of digital filters: finite impulse response (FIR) filters infinite (IIR) filters. FIR filter can be guaranteed to have linear phase always stable filters, so is widely applicable. The differential evolution (DE) algorithm, which proposed particularly for numeric optimization problems, a population-based algorithm like the genetic algorithms. In this work, new DE based on reserved...
Semi-supervised clustering algorithms have several limitations: 1) the computation complexity of them is very high, because calculating similarity distances pairs examples time-consuming; 2) traditional semi-supervised methods not considered how to make full use must-link and cannot-link constraints. In clustering, contribution a few pairwise constraints performance limited, some may negatively affect outcome; 3) these are effective handle high dimensional data, especially for time series...
Traditional time series classification problem with supervised learning algorithm needs a large set of labeled training data. In reality, the number data is often smaller and there huge unlabeled However, manually labeling these examples time-consuming expensive, sometimes it even impossible. Although some semi-supervised active methods were proposed to handle univariate data, few work have touched positive for multivariate (MTS) due being more complex. this paper we focus on First, propose...
Multivariate time series (MTS) classification is an important topic in data mining, and lots of efficient models techniques have been introduced to cope with it. However, early on imbalanced MTS largely remains open problem. To deal this issue, we adopt a multiple under-sampling dynamical subspace generation method obtain initial training data, each used learn base learner. Finally, ensemble classifier for data. Experimental results show that our proposed methods can achieve effective prediction
Recently, multivariate time series (MTS) clustering has gained lots of attention. However, state-of-the-art algorithms suffer from two major issues. First, few existing studies consider correlations and redundancies between variables MTS data. Second, since different clusters usually exist in intrinsic variables, how to efficiently enhance the performance by mining a cluster is challenging work. To deal with these issues, we first propose variable-weighted K-medoids algorithm (VWKM) based on...