A5100614090- Optical measurement and interference techniques
- Supercapacitor Materials and Fabrication
- Advanced Vision and Imaging
- Structural Health Monitoring Techniques
- Advanced Image and Video Retrieval Techniques
- Adaptive Control of Nonlinear Systems
- Video Surveillance and Tracking Methods
- Diamond and Carbon-based Materials Research
- Metal and Thin Film Mechanics
- Electrocatalysts for Energy Conversion
- Structural Engineering and Vibration Analysis
- Advanced Neural Network Applications
- Robotics and Sensor-Based Localization
- Gas Sensing Nanomaterials and Sensors
- Optical Systems and Laser Technology
- Advanced battery technologies research
- Concrete Corrosion and Durability
- Analytical Chemistry and Sensors
- Advanced Measurement and Detection Methods
- Vibration Control and Rheological Fluids
- Advancements in Battery Materials
- Smart Materials for Construction
- 3D Surveying and Cultural Heritage
- Transition Metal Oxide Nanomaterials
- Advanced Sensor and Energy Harvesting Materials
North University of China
2022-2025
Criminal Investigation Police University of China
2011-2024
University of Science and Technology Beijing
2023-2024
China University of Mining and Technology
2023-2024
Beihang University
2010-2024
Harbin Engineering University
2017-2024
Hong Kong R&D Centre for Logistics and Supply Chain Management Enabling Technologies
2024
Baidu (China)
2022-2024
FirstUnion (China)
2024
University of Central Florida
2020-2024
The innovative core-ring structured NiCo2O4 nanoplatelets were found to be novel and promising photocatalysts. physical photophysical properties of the photocatalyst characterized by SEM, TEM, XPS, UV−vis absorption, photoluminescence, respectively. composed much smaller nanocrystallines, with an average size 80−150 nm, compared ordinary prepared through a conventional hydroxide decomposition method. Moreover, optical band gap energies estimated 2.06 3.63 eV from absorption spectra....
The pursuit of harmonic combination technology and fashion intrinsically points to the development smart garments. Herein, we present an all-solid tailorable energy textile possessing integrated function simultaneous solar harvesting storage, call it device. Our technique makes possible tailor multifunctional into any designed shape without impairing its performance produce stylish garments for wearable self-powering system with enhanced user experience more room design. "threads" (fiber...
Summary This article investigates the composite adaptive fuzzy finite‐time prescribed performance control issue of switched nonlinear systems subject to unknown external disturbance and requirement. First, by utilizing compensation prediction errors, piecewise parameter update law is employed improve approximation accuracy nonlinearity. Then, improved fractional‐order filter error signal are introduced cope with influences caused explosive calculation error, respectively. Meanwhile, effect...
A facile two-step method has been developed to synthesize 3D core/shell-structured composites (CNTs@Ni–Co–S) composed of ternary nickel cobalt sulfide nanosheets (Ni–Co–S) as the shell and carbon nanotubes (CNTs) core on a flexible Ni mesh (Ni@CNTs@Ni–Co–S).
Modern GPU datacenters are critical for delivering Deep Learning (DL) models and services in both the research community industry. When operating a datacenter, optimization of resource scheduling management can bring significant financial benefits. Achieving this goal requires deep understanding job features user behaviors. We present comprehensive study about characteristics DL jobs management. First, we perform large-scale analysis real-world traces from SenseTime. uncover some interesting...
Federated learning (FL) is a privacy-preserving distributed machine paradigm that enables multiple clients to collaboratively train statistical models without disclosing raw training data. However, the inaccessible local data and uninspectable process make FL susceptible various Byzantine attacks (e.g., poisoning model attacks), aiming manipulate degrade performance. Most of existing Byzantine-robust schemes cannot effectively defend against stealthy craft poisoned statistically similar...
The performance of video frame interpolation is inherently correlated with the ability to handle motion in input scene. Even though previous works recognize utility asynchronous event information for this task, they ignore fact that may or not result blur be interpolated, depending on length exposure time frames and speed motion, assume either sharp, restricting themselves interpolation, it blurry, including an explicit, separate deblurring stage before their pipeline. We instead propose a...
Amorphous NiSe was deposited on the carbon nanotubes (CNTs) networks by pulsed laser deposition method forming 2 -1 -1CC@CNT@NiSe composites.The CC
In this work we propose a new automatic image annotation model, dubbed diverse and distinct (D2IA). The generative model D2IA is inspired by the ensemble of human annotations, which create semantically relevant, yet tags. D2IA, generate relevant tag subset, in tags are to contents each other, using sequential sampling from determinantal point process (DPP) model. Multiple such subsets that cover semantic aspects or levels generated randomly perturbing DPP process. We leverage adversarial...
The COVID-19 pandemic is an evolving situation in the United States and spreading at alarming rates. adoption of public health-informed hygienic practices can have a large impact on community transmission including wearing face masks settings. Convolutional Neural Networks (CNN) be trained to classify people with impressive accuracy. However, current mask datasets contain clear, high-resolution close-up images individuals which unrepresentative lower fidelity distant faces more prominent...
Pre-trained language models have achieved state-of-the-art results in various Natural Language Processing (NLP) tasks. GPT-3 has shown that scaling up pre-trained can further exploit their enormous potential. A unified framework named ERNIE 3.0 was recently proposed for pre-training large-scale knowledge enhanced and trained a model with 10 billion parameters. outperformed the on NLP In order to explore performance of 3.0, we train hundred-billion-parameter called Titan 260 parameters...