A5101825362- Extraction and Separation Processes
- Geochemistry and Elemental Analysis
- Algal biology and biofuel production
- Adsorption and biosorption for pollutant removal
- Metabolomics and Mass Spectrometry Studies
- Fault Detection and Control Systems
- Chromium effects and bioremediation
- Image and Object Detection Techniques
- Industrial Vision Systems and Defect Detection
- Manufacturing Process and Optimization
- Microbial Community Ecology and Physiology
- Plant-Microbe Interactions and Immunity
- Adaptive Control of Nonlinear Systems
- Iterative Learning Control Systems
- Environmental Toxicology and Ecotoxicology
- Business Strategy and Innovation
- Machine Fault Diagnosis Techniques
- Electric Power Systems and Control
- Non-Destructive Testing Techniques
- Inertial Sensor and Navigation
- Control and Dynamics of Mobile Robots
- Analytical chemistry methods development
- Advanced ceramic materials synthesis
- Advanced Neural Network Applications
- Neural Networks and Reservoir Computing
Central South University
2022-2024
University of Southern California
2024
Zhejiang University of Technology
2024
RWTH Aachen University
2023
State Key Laboratory of High Performance Complex Manufacturing
2022
Harbin Institute of Technology
2006
Surface defect detection of printed circuit boards (PCBs) is a critical stage in ensuring product quality on production lines electronics manufacturing. The excellent performance methods using deep learning models such as convolutional neural networks (CNNs) and autoencoders limited by image uncertainty under uneven ambient light or unstable transmission channels label due to human perception errors lack expert knowledge. To overcome these difficulties, novel collaborative classification...
Lately, the practice of utilizing task-specific fine-tuning has been implemented to improve performance large language models (LLM) in subsequent tasks. Through integration diverse LLMs, overall competency LLMs is significantly boosted. Nevertheless, traditional ensemble methods are notably memory-intensive, necessitating simultaneous loading all specialized into GPU memory. To address inefficiency, model merging strategies have emerged, one reduce memory footprint during inference. Despite...
HIT spherical robot has been designed on the principle of carrying out turning and driving motions independently in order to reduce dynamic complexity realize real-time detection. After introducing decoupling implemented working processing, we derive equations about two fundamental motions, i.e., rolling turning. Hardware software open-loop control system are introduced briefly and, following, experiments show that can fulfill common exploration applying proposed method.
Ensuring safety for vehicle overtaking systems is one of the most fundamental and challenging tasks in autonomous driving. This task particularly intricate when must not only overtake its front safely but also consider presence potential opposing vehicles opposite lane that it will temporarily occupy. In order to tackle such scenarios, we introduce a novel integrated framework tailored manoeuvres. Our approach integrates theories varying-level control barrier functions (CBF) time-optimal...
Large pre-trained models (LPMs), such as LLaMA and GLM, have shown exceptional performance across various tasks through fine-tuning. Although low-rank adaption (LoRA) has emerged to cheaply fine-tune these LPMs on downstream tasks, their deployment is still hindered by the vast model scale computational costs. Neural network pruning offers a way compress LPMs. However, current methods designed for are not compatible with LoRA. This due utilization of unstructured LPMs, impeding merging LoRA...
Lately, the practice of utilizing task-specific fine-tuning has been implemented to improve performance large language models (LLM) in subsequent tasks. Through integration diverse LLMs, overall competency LLMs is significantly boosted. Nevertheless, traditional ensemble methods are notably memory-intensive, necessitating simultaneous loading all specialized into GPU memory. To address inefficiency, model merging strategies have emerged, one reduce memory footprint during inference. Despite...
To solve the coupling problem of control loops and adaptive parameter tuning in multi-input multi-output (MIMO) PID system, a self-adaptive LSAC-PID algorithm is proposed based on deep reinforcement learning (RL) Lyapunov-based reward shaping this paper. For complex unknown mobile robot environment, an RL-based MIMO hybrid strategy firstly presented. According to dynamic information environmental feedback robot, RL agent can output optimal parameters real time, without knowing mathematical...
The electric drivetrain of an autonomous driving vehicle is not only supposed to be efficient, but there are also high requirements for reliability and fault-tolerant capability. Inter-turn short circuits (ITSCs) can occur frequently have a severe impact on the drivetrain. In this work, machine learning method based convolutional neural networks (CNNs) detects fault in binary classification problem identifies eleven different classes multi-class classification. faulty machine's...