A5101864810- Natural Language Processing Techniques
- Topic Modeling
- Recommender Systems and Techniques
- Adsorption and biosorption for pollutant removal
- Water Quality Monitoring and Analysis
- Privacy-Preserving Technologies in Data
- Adversarial Robustness in Machine Learning
- Water Quality Monitoring Technologies
- Advanced oxidation water treatment
- Advanced Photocatalysis Techniques
- Robotic Path Planning Algorithms
- Nanomaterials for catalytic reactions
- Advanced Text Analysis Techniques
- Simulation and Modeling Applications
- Zeolite Catalysis and Synthesis
- Gallbladder and Bile Duct Disorders
- Maritime Transport Emissions and Efficiency
- Catalytic Processes in Materials Science
- Pediatric Hepatobiliary Diseases and Treatments
- Text and Document Classification Technologies
- Pancreatic and Hepatic Oncology Research
- Robotics and Sensor-Based Localization
- Membrane-based Ion Separation Techniques
- Aerogels and thermal insulation
- Image and Video Quality Assessment
Northeast Forestry University
2024-2025
China University of Petroleum, Beijing
2025
Capital Institute of Pediatrics
2025
Dalian Maritime University
2024
Ministry of Education of the People's Republic of China
2024
Jilin Province Science and Technology Department
2024
Jilin University
2024
Kunming University of Science and Technology
2023
Tongji University
2007-2021
Shandong University of Technology
2006-2020
Selective conversion of ethanol to high yields ethylene under mild conditions.
Mobility analysis is a crucial element in the research area of transportation systems. Forecasting traffic information offers viable solution to address conflict between increasing demands and limitations infrastructure. Predicting human travel significant aiding various urban management tasks, such as taxi dispatch planning. Machine learning deep methods are favored for their flexibility accuracy. Nowadays, with advent large language models (LLMs), many researchers have combined these...
Early diagnosis of biliary atresia (BA) is crucial for improving patient outcomes, yet remains a significant global challenge. This challenge may be ameliorated through the application artificial intelligence (AI). Despite promise AI in medical diagnostics, its to multimodal BA data has not achieved substantial breakthroughs. study aims leverage diverse sources and formats develop an intelligent diagnostic system BA. We constructed largest known dataset, comprising ultrasound images,...
With the rapid advancement of artificial intelligence technology, usage machine learning models is gradually becoming part our daily lives. High-quality rely not only on efficient optimization algorithms but also training and processes built upon vast amounts data computational power. However, in practice, due to various challenges such as limited resources privacy concerns, users need often cannot train locally. This has led them explore alternative approaches outsourced federated learning....
With the rapid growth of multimodal data in social medias and huge requirement short but abundant information. Multimodal summarization has drawn much attention both industry academia. It usually obtains textual summary from multiple sources by computer vision or nature language processing technologies. However, there are also two challenges modeling such task: 1) The feature representation is limited non-alignment among data; 2) Massive parallel required during training, which...
Reinforcement learning (RL) is a technique based on trial and error. Q-learning method of RL algorithms. It has been applied widely in the adaptive path planning for autonomous mobile robot. In order to decrease space increase convergent speed, this paper adopts Q-layered divide task searching optimal into three basic behaviors (or subtasks), namely static obstacle-avoidance, dynamic obstacle-avoidance goal approaching. Especially behavior, novel priority Q search (PQA) used avoid blindly...
Internet users are benefiting from technologies of abstractive summarization enabling them to view articles on the internet by reading article summaries only instead an entire article. However, there disadvantages for analyzing with texts and images due semantic gap between vision language. These focus more aggregating features neglect heterogeneity each modality. At same time, lack consideration intrinsic data properties within modality information cross-modal correlations result in poor...
With the recent success of large language models, particularly foundation models with generalization abilities, applying for recommendations becomes a new paradigm to improve existing recommendation systems. It open challenge enable model capture user preference changes in timely manner reasonable communication and computation costs while preserving privacy. This paper proposes novel federated adaptation mechanism enhance model-based system privacy-preserving manner. Specifically, each...
Federated learning allows several clients to train one machine model jointly without sharing private data, providing privacy protection. However, traditional federated is vulnerable poisoning attacks, which can not only decrease the performance, but also implant malicious backdoors. In addition, direct submission of local parameters lead leakage training dataset. this paper, we aim build a privacy-preserving and Byzantine-robust scheme provide an environment with no vandalism (NoV) against...
With the recent success of large language models, particularly foundation models with generalization abilities, applying for recommendations becomes a new paradigm to improve existing recommendation systems. It open challenge enable model capture user preference changes in timely manner reasonable communication and computation costs while preserving privacy. This paper proposes novel federated adaptation mechanism enhance model-based system privacy-preserving manner. Specifically, each...
The leakage phenomenon has become a major challenge in engineering applications for vacuum adsorption technology. To achieve efficient design optimization of the system based on finite element model, an method residual adaptive fitting surrogate model is proposed. performance and influencing parameters established under conditions by response surface (RSM). residuals generated RSM are fitted using kriging incorporated into construction RSM. models updated learning functions, combining...
In this paper, a model-based strategy is used to realize rapid searching of the shortest path in order solve global optimizing problem planning mobile robot. The Hough transform build cell map working environment for robot by extracting fringes information solid obstacles and them into 2D plane. We simplify construction state space design search based on free cells' evaluation distance which between start point goal point. relatively path's realized through tree. experimental simulation...
Deep learning has become a key technology on modeling large amounts of multi-sourced data. For privacy concerns, the data sharing among companies and organizations is increasingly difficult. In this paper, we present crowd-sourced federated solution to train neural networks with hybrid blockchain architecture. Smart contracts are used share authentications main chain, where proxy re-encryption for preserving. A consensus-based asynchronous practical byzantine (APBFL) algorithm proposed side...
In order to deal with the nonlinear structure and time-varying characteristics of intelligent vehicle (IV) systems, this paper presents an idea pattern-space-cover construct IV lateral motion model. The construction method is deduced based on fuzzy C means (FCM) algorithm can validate cluster sampled data set. Cluster domains are clear, subspaces partitioned distinctly. It provides a theoretical basis for pattern-space. simulation results show that effective feasible.