A5064619395- Recommender Systems and Techniques
- Data Management and Algorithms
- Caching and Content Delivery
- Advanced Graph Neural Networks
- Graph Theory and Algorithms
- Human Mobility and Location-Based Analysis
- Machine Fault Diagnosis Techniques
- Advanced Image and Video Retrieval Techniques
- Domain Adaptation and Few-Shot Learning
- Optical Systems and Laser Technology
- Retinal Imaging and Analysis
- Gear and Bearing Dynamics Analysis
- Electrical and Bioimpedance Tomography
- Civil and Geotechnical Engineering Research
- Text and Document Classification Technologies
- Brain Tumor Detection and Classification
- Advanced Optical Sensing Technologies
- Retinal and Optic Conditions
- Advanced Neural Network Applications
- Structural Integrity and Reliability Analysis
- Machine Learning and Data Classification
- Web Data Mining and Analysis
- Astronomical Observations and Instrumentation
- Network Security and Intrusion Detection
- Stock Market Forecasting Methods
Ministry of Agriculture and Rural Affairs
2022-2025
Shandong Agricultural University
2022-2025
Nanjing University
2010-2024
University of Shanghai for Science and Technology
2024
Cisco College
2024
Commercial Aircraft Corporation of China (China)
2021-2023
Tianjin University of Technology
2015-2022
Henan Normal University
2006-2021
Kaifeng University
2019
UNSW Sydney
2010-2016
CC-Cruiser is an artificial intelligence (AI) platform developed for diagnosing childhood cataracts and providing risk stratification treatment recommendations. The high accuracy of was previously validated using specific datasets. objective this study to compare the diagnostic efficacy decision-making capacity between ophthalmologists in real-world clinical settings.This multicentre randomized controlled trial performed five ophthalmic clinics different areas across China. Pediatric...
Multi-label image recognition is a challenging computer vision task of practical use. Progresses in this area, how-ever, are often characterized by complicated methods, heavy computations, and lack intuitive explanations. To effectively capture different spatial regions occupied objects from categories, we propose an embarrassingly simple module, named class-specific residual attention (CSRA). CSRA generates features for every category proposing score, then combines it with the...
Purpose To establish and validate a universal artificial intelligence (AI) platform for collaborative management of cataracts involving multilevel clinical scenarios explored an AI-based medical referral pattern to improve efficiency resource coverage. Methods The training validation datasets were derived from the Chinese Medical Alliance Artificial Intelligence, covering healthcare facilities capture modes. labelled using three-step strategy: (1) mode recognition; (2) cataract diagnosis as...
Given a query graph $q$ and data G, computing all occurrences of q in namely exact all-matching, is fundamental analysis with wide spectrum real applications. It challenging since even finding one occurrence G (subgraph isomorphism test) NP-Complete. Consider that many applications, exploratory queries from users are often inaccurate to express their demands. In this paper, we study the problem efficiently approximate G. Particularly, retrieving matches number possible missing edges bounded...
With the increasing popularity of Location-Based Social Networks (LBSNs), a significant volume check-in data users has been generated. Such massive brings difficulties for to efficiently retrieve their desired point-of-interest (POI). As result, POI recommendation systems have received extensive attention from academia and industry. Currently, most existing approaches only provide with fixed set recommended POIs based on historical records users, cannot achieve flexible feasible...
The grounding grid is an important piece of equipment to ensure the safety a power system, and thus research detecting on its corrosion status great significance. Electrical impedance tomography (EIT) effective method for imaging. However, inverse process image reconstruction has pathological solutions, which lead unstable imaging results. This paper proposes electrical based improved conditional generative adversarial network (CGAN), aiming improve precision accuracy. Its generator combines...
In order to ensure the price stability of niche agricultural products and enhance farmers’ income, study delves into pattern ginger fluctuation rule its main influencing factors. By combining seasonal decomposition STL, long short-term memory network LSTM, attention mechanism ATT Kolmogorov-Arnold network, a combined STL-LSTM-ATT-KAN prediction model is developed, parameters are finely tuned by using multi-population adaptive particle swarm optimisation algorithm (AMP-PSO). Based on an...
This study focuses on the landslide susceptibility assessment in Maerkang City, Sichuan Province. Twelve evaluation factors, including terrain relief and slope, were selected for analysis. The employs Information Value-Analytic Hierarchy Process (IV-AHP), Random Forest (RF), Extreme Gradient Boosting (XGBoost), as well hybrid models IV-RF IV-XGBoost to evaluate explore differences between traditional methods, machine learning models, models. findings indicate that statistical analysis...
In finetuning a large pretrained model to downstream tasks, parameter-efficient fine-tuning (PEFT) methods can effectively finetune models with few trainable parameters, but suffer from high GPU memory consumption and slow training speed. Because learnable parameters these are entangled the model, gradients related frozen model's have be computed stored during finetuning. We propose Low-rank Attention Side-Tuning (LAST), which disentangles module by freezing not only also outputs of network....
When pre-trained models become rapidly larger, the cost of fine-tuning on downstream tasks steadily increases, too. To economically fine-tune these models, parameter-efficient transfer learning (PETL) is proposed, which only tunes a tiny subset trainable parameters to efficiently learn quality representations. However, current PETL methods are facing dilemma that during training GPU memory footprint not effectively reduced as parameters. will likely fail, too, if full encounters...
A supergraph containment search is to retrieve the data graphs contained by a query graph. In this paper, we study problem of efficiently retrieving all approximately graph, namely similarity on containment. We propose novel and efficient index boost efficiency processing. have studied processing cost two construction strategies aimed at optimizing performance different types graphs: top-down strategy bottom-up strategy. Moreover, indexing technique proposed effectively merging indexes...
The accurate prediction of scallion prices can not only optimize supply chain management and help related practitioners consumers to make more reasonable purchasing decisions, but also provide guidance for farmers’ planting choices, thus enhancing market efficiency promoting the sustainable development whole industry. This study adopts idea decomposition–denoising–aggregation, using three decomposition denoising techniques combined with single models form a base model. Various are divided...
The frequent and sharp fluctuations in garlic prices seriously affect the sustainable development of industry. Accurate prediction can facilitate correct evaluation scientific decision making by practitioners, thereby avoiding market risks promoting healthy To improve accuracy prices, this paper proposes a garlic-price-prediction method based on combination long short-term memory (LSTM) multiple generalized autoregressive conditional heteroskedasticity (GARCH)-family models for nonstationary...
Self-supervised learning (SSL) methods targeting scene images have seen a rapid growth recently, and they mostly rely on either dedicated dense matching mechanism or costly unsupervised object discovery module. This paper shows that instead of hinging these strenuous operations, quality image representations can be learned by treating scene/multi-label SSL simply as multi-label classification problem, which greatly simplifies the framework. Specifically, multiple binary pseudo-labels are...
Neural network quantization aims to accelerate and trim full-precision neural models by using low bit approximations. Methods adopting the aware training (QAT) paradigm have recently seen a rapid growth, but are often conceptually complicated. This paper proposes novel highly effective QAT method, quantized feature distillation (QFD). QFD first trains (or binarized) representation as teacher, then quantize knowledge (KD). Quantitative results show that is more flexible (i.e., friendly) than...
Given a query graph <inline-formula><tex-math notation="LaTeX">$q$</tex-math></inline-formula> , retrieving the data graphs notation="LaTeX">$g$</tex-math></inline-formula> from set <inline-formula> <tex-math notation="LaTeX">$D$</tex-math></inline-formula> of such that notation="LaTeX">$q$</tex-math> </inline-formula> contains namely supergraph containment search, is fundamental in analysis with wide range real applications. It very challenging due to NP-Completeness subgraph isomorphism...