A5100758591- Heat Transfer and Boiling Studies
- Web Data Mining and Analysis
- Visual Attention and Saliency Detection
- Heat Transfer and Optimization
- Metaheuristic Optimization Algorithms Research
- Fluid Dynamics and Mixing
- Coding theory and cryptography
- Advanced Image and Video Retrieval Techniques
- Solar Thermal and Photovoltaic Systems
- Gas Dynamics and Kinetic Theory
- Remote-Sensing Image Classification
- Advanced Image Processing Techniques
- Advanced Sensor and Control Systems
- Software-Defined Networks and 5G
- Advanced Image Fusion Techniques
- Peer-to-Peer Network Technologies
- Complex Network Analysis Techniques
- Optimization and Packing Problems
- Advanced Vision and Imaging
- Caching and Content Delivery
- Building Energy and Comfort Optimization
- Advanced Optical Network Technologies
- Advanced Algorithms and Applications
- Spacecraft and Cryogenic Technologies
- Advanced Neural Network Applications
North China Electric Power University
2007-2024
Hunan University
2021-2024
Anhui University
2024
Tsinghua University
2022
Xiamen University
2019-2021
Xidian University
2020
China Academy of Space Technology
2020
Electric Power University
2020
Zhejiang University
2020
Nanjing University of Aeronautics and Astronautics
2019
Transformers have achieved remarkable results in single-image super-resolution (SR). However, the challenge of balancing model performance and complexity has hindered their application lightweight SR (LSR). To tackle this challenge, we propose an efficient striped window transformer (ESWT). We revisit normalization layer design a concise structure to build ESWT. Furthermore, introduce mechanism long-term dependencies more efficiently. fully exploit potential ESWT, novel flexible training...
Salient object detection is a binary pixel-wise classification to distinguish objects in an image, and also have attracted many research interests the optical Remote Sensing Images (RSIs). The existing state-of-the-art method exploits full encoder-decoder architecture predict salient RSIs, suffering from problem of unsmooth edges incomplete structures. To address these problems, this paper, we propose Boundary-Aware Network (BANet) with two-stage partial decoders sharing same encoders for...
Unit equipment is the key to industrial production, and predicting unit failures focus of improving productivity. In order improve accuracy reliability predicting, consider use multiple related influencing factors for prediction. This paper presents a fault prediction method based on CNN-LSTM. Firstly, data variables that affects predicted value over period time are formed into large matrix. Then, Long short-term memory network trained by feature information extracted from convolutional...
The classical person re-identification methods are mostly focused on employing discriminative features amongst which the distance is measured Euclidean space, while effort of re-ranking constrained as lack utilization quality context representation in embedding set.In this paper, we incorporate graph models feature subsets resorting to initial ranking by adopting integration attention mechanism into convolution network.On one hand, information regarding pairs considered compute group...
In order to solve some existing problems in traditional data preprocessing technology for web log mining, an improved is used this article. The identification strategy based on the referred page adopted at stage of user identification, which more effective than one site topology. At Session Identification, fixed priori threshold combined with session reconstruction introduced. First, initial set developed by method threshold, and then optimized using reconstruction. Experiments have proved...
Based on the combination of five user minimum browsing behaviors, this paper focuses a comprehensive analysis quantitative relations between time and interest rate, proposed new method calculating degree thematic interest, is based association rule. Obtain user's by fuzzy clustering algorithm, establish model. The experiments confirm that rate reasonable effective, model responses exactly, have been effectively used in personalized service.
Abstract Objective: In order to cope with a sudden outbreak of new coronavirus infection, large number potential infected persons need be isolated. A smart monitoring system which integrates Internet things and blockchain technology monitor isolated people in real time was design studied.Methods: internet devices will collects the location physical data people, sent master integrate format those transfer contract. contract compares analyses threshold is predefined. When exceed threshold,...
In order to improve the reliability and accuracy of prediction results in rotating machinery fault trend, this paper proposes a method based on Bi-LSTM Attention mechanism. Firstly, is constructed extract timing signal data characteristics machinery. Then, mechanism introduced highlight sequence that has key influence result. Finally, difference between predicted value model outputs empirical threshold compared establish multi-step, multivariable model. The actual field used train built...
Two dimensional strip packing problem implies a sort of NP-hard combinational optimization, where set rectangles are defined all for filling with finite width and infinite height being minimized. Optimization taking no account rotation guillotine constraints remains problematic in solving the problem. This paper proposes new variant hybrid heuristic algorithm optimizing solutions involves three key improvements: building δ scoring rules selecting rectangles, use red-black trees that stores...
Link selection algorithm is an important part in a vertical search engine. Better leads to better In order build nice engine, we propose new kind of link algorithm, based on the famous HITS. The proposed aims collect high topic relevance pages specific domain. There are two factors this one expanded meta-data score, which calculated by combining analysis content with hyperlink structure characteristics; other inherit influence father pages' score authority value and hub value. Experiment...
This paper proposes a self-organizing control system for uncertain nonlinear systems. The proposed neural network is composed of conventional brain emotional learning (BEL) and cerebellar model articulation controller (CMAC). input value the feed to BEL channel CMAC channel. output generated by comprehensive action two channels. structure dynamic, using algorithm allows increasing or decreasing weight layers. parameters are on-line tuned rules; updating rules robust derived from Lyapunov...