A5100671185- Model Reduction and Neural Networks
- Markov Chains and Monte Carlo Methods
- Plant Micronutrient Interactions and Effects
- Gaussian Processes and Bayesian Inference
- Advanced Fiber Laser Technologies
- Aluminum toxicity and tolerance in plants and animals
- Advanced Frequency and Time Standards
- Photonic and Optical Devices
- Plant Stress Responses and Tolerance
- Numerical methods for differential equations
- Seismic Imaging and Inversion Techniques
- Advanced Mathematical Modeling in Engineering
- Cold Atom Physics and Bose-Einstein Condensates
- Stochastic Gradient Optimization Techniques
- Micro and Nano Robotics
- Target Tracking and Data Fusion in Sensor Networks
- Rough Sets and Fuzzy Logic
- X-ray Spectroscopy and Fluorescence Analysis
- Diffusion and Search Dynamics
- Sentiment Analysis and Opinion Mining
- Mathematical Approximation and Integration
- Advanced Numerical Methods in Computational Mathematics
- Trace Elements in Health
- Particle accelerators and beam dynamics
- Microwave Imaging and Scattering Analysis
Xi'an Jiaotong University
2022-2023
University of Chinese Academy of Sciences
2022-2023
Chinese Academy of Sciences
2010-2023
Hunan Institute of Technology
2023
Shanghai Institute of Optics and Fine Mechanics
2021-2023
Civil Aviation University of China
2023
Beijing Normal University
2018-2022
University of Hong Kong
2022
Wuhan Ship Development & Design Institute
2022
Anhui Medical University
2022
Large areas of soil in southern China are contaminated with cadmium (Cd) and deficient boron (B). Previously, we suggested that B supplementation could reduce Cd accumulation hot peppers (Capsicum annuum L.); however, the physiological mechanisms underlying this reduction remain unclear. In study, uptake translocation pepper plants were investigated using hydroponic experiments different treatments. A pot experiment was performed to verify whether decreased concentration by minimizing rate....
The objective of this paper is to design novel multi-layer neural network architectures for multiscale simulations flows taking into account the observed data and physical modeling concepts. Our approaches use deep learning concepts combined with local model reduction methodologies predict flow dynamics. Using reduced-order important constructing robust since models provide fewer degrees freedom. Flow dynamics can be thought as networks. More precisely, solution (e.g., pressures saturations)...
We demonstrate an ultrastable miniaturized transportable laser system at 1550 nm by locking it to optical fiber delay line (FDL). To achieve optimized long-term frequency stability, the FDL was placed into a vacuum chamber with five-layer thermal shield, and delicate two-stage active temperature stabilization, power RF stabilization were applied in system. A fractional stability of better than 3.2×10-15 1 s averaging time 1.1×10-14 1000 achieved, which is best all-fiber-based observed date.
Uncertainty quantification (UQ) of reservoir heterogeneity is essential to accurately infer fluid flow behavior in subsurface formations, and the task often accomplished by integrating forward physics simulators with iterative data assimilation methods, such workflows are usually computationally expensive due nature prohibitive cost simulations.In this work, we develop a new Ensemble Multi-Fidelity Neural Network(EMF-Net) mitigate efficiency bottleneck UQ. EMF-Net directly infers uncertain...
We demonstrate an all-fiber-based photonic microwave generation with 10-15 frequency instability. The system consists of ultra-stable laser by optical fiber delay line, “figure-of-nine” comb, a high signal-to-noise ratio detection unit, and synthesizer. whole links are made from components, which renders the compactness, reliability, robustness respect to environmental influences. Frequency instabilities 3.5×10-15 at 100 s for 6.834 GHz signal 4.3×10-15 9.192 were achieved.
Physics-informed neural network (PINN) has been successfully applied in solving a variety of nonlinear non-convex forward and inverse problems. However, the training is challenging because loss functions multiple optima Bayesian problem. In this work, we propose multi-variance replica exchange stochastic gradient Langevin diffusion method to tackle challenge local optimization modal posterior distribution Replica methods are capable escaping from traps accelerating convergence. it may not be...
In this paper, we combine deep learning concepts and some proper orthogonal decomposition (POD) model reduction methods for predicting flow in heterogeneous porous media. Nonlinear dynamics is studied, where the regarded as a multi-layer network. The solution at current time step network of initial input parameters. As input, consider various sources, which include source terms (well rates), permeability fields, conditions. We dynamics, known locations data integrated to by modifying...
We demonstrate a temperature-insensitive fiber-delay-line-stabilized (FDL-stabilized) laser based on dual Mach-Zehnder interferometer (MZI) by using polarization maintaining fibers (PMFs). Two orthogonal components of beam are simultaneously transmitted in the interferometer. Each component exhibits unique phase shift response to changes temperature, forming MZI. One heterodyne signals is used lock frequency, while other one compensate frequency change induced temperature fluctuation. The...
Due to the advantages of small, light and low-cost, Microelectromechanical system (MEMS) gyroscopes are used in area navigation. To improve accuracy MEMS gyroscopes, an outfield calibration scheme is proposed. The bias, scale errors installation can be estimated simultaneously without high precision angular velocity reference provided by rotation platform. Firstly, relation between gyroscope output accelerometer built. Secondly, according 12-parameter error model gyroscope, obtained....
We present a highly stable and miniaturised optical system designed for caesium satellite-borne cold atom clock (CSCAC). The was integrated on <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$224\,{\mathop{\rm mm}\nolimits} \times 195\,{\mathop{\rm 112\,{\mathop{\rm $</tex-math></inline-formula> aluminium-based silicon carbide bench emitted laser beams with four different frequencies three fiber ports to...
Physics-informed neural network (PINN) has been successfully applied in solving a variety of nonlinear non-convex forward and inverse problems. However, the training is challenging because loss functions multiple optima Bayesian problem. In this work, we propose multi-variance replica exchange stochastic gradient Langevin dynamics method to tackle challenge local optimization modal posterior distribution Replica methods are capable escaping from traps accelerating convergence; two chains...
Robust control design for quantum systems is a challenging and key task practical technology. In this work, we apply neural networks to learn the problem semiclassical Schrödinger equation, where variable potential given by an external field that may contain uncertainties. Inspired relevant work [29], incorporate sampling-based learning process into training of networks, while combining with fast time-splitting spectral method equation in regime. The numerical results have shown efficiency...
In the present work, mechanism of O 2 ( 1 Δ g ) generation from reaction dissolved Cl with H in basic aqueous solution has been explored by combined ab initio calculation and nonadiabatic dynamics simulation, together different solvent models. Three possible pathways have determined for generation, but two them are sequentially downhill processes until formation OOCl − complex water, which high exothermic character. Once is formed, singlet molecular oxygen easily generated its decomposition...
In this work, we propose a network which can utilize computational cheap low-fidelity data together with limited high-fidelity to train surrogate models, where the multi-fidelity are generated from multiple underlying models. The takes context set as input (physical observation points, low fidelity solution at observed points) and output (high pairs. It uses neural process learn distribution over functions conditioned on sets provide mean standard deviation target sets. Moreover, proposed...
Significant evidence is available to support the quantum effects of gravity that leads generalized uncertainty principle (GUP) and minimum observable length. Usually mechanics, statistical physics doesn't take into account. Thermodynamic properties ideal Bose gases in different external power-law potentials are studied under GUP with physical method. Critical temperature, internal energy, heat capacity, entropy, particles number ground state excited calculated analytically GUP. Below...