A5052016306- Model Reduction and Neural Networks
- Neural Networks and Applications
- Machine Learning and Data Classification
- Fluid Dynamics and Turbulent Flows
- Advancements in Photolithography Techniques
- Quantum Mechanics and Applications
- Explainable Artificial Intelligence (XAI)
- Computational Physics and Python Applications
- Quantum Computing Algorithms and Architecture
- Reservoir Engineering and Simulation Methods
- Meteorological Phenomena and Simulations
- Machine Learning and Algorithms
- Quantum Information and Cryptography
- Fluid Dynamics and Vibration Analysis
- Advanced Mathematical Modeling in Engineering
- CO2 Sequestration and Geologic Interactions
- Machine Learning in Materials Science
- Nuclear Engineering Thermal-Hydraulics
- Image and Signal Denoising Methods
- Generative Adversarial Networks and Image Synthesis
- Homotopy and Cohomology in Algebraic Topology
- Advanced Algebra and Geometry
- Advanced Electron Microscopy Techniques and Applications
- Advanced Neural Network Applications
- CCD and CMOS Imaging Sensors
Huazhong University of Science and Technology
2024-2025
Shanghai Jiao Tong University
2024
California Institute of Technology
1989-2024
Heilongjiang Institute of Technology
2024
Harbin Institute of Technology
2024
Soochow University
2023-2024
Northeast Forestry University
2024
Shantou University
2023-2024
University of Liverpool
2023
China University of Geosciences (Beijing)
2023
The classical development of neural networks has primarily focused on learning mappings between finite-dimensional Euclidean spaces. Recently, this been generalized to operators that learn function For partial differential equations (PDEs), directly the mapping from any functional parametric dependence solution. Thus, they an entire family PDEs, in contrast methods which solve one instance equation. In work, we formulate a new operator by parameterizing integral kernel Fourier space,...
FourCastNet, short for Fourier Forecasting Neural Network, is a global data-driven weather forecasting model that provides accurate to medium-range predictions at $0.25^{\circ}$ resolution. FourCastNet accurately forecasts high-resolution, fast-timescale variables such as the surface wind speed, precipitation, and atmospheric water vapor. It has important implications planning energy resources, predicting extreme events tropical cyclones, extra-tropical rivers. matches accuracy of ECMWF...
The classical development of neural networks has been primarily for mappings between a finite-dimensional Euclidean space and set classes, or two spaces. purpose this work is to generalize so that they can learn infinite-dimensional spaces (operators). key innovation in our single network parameters, within carefully designed architecture, may be used describe different approximations those We formulate approximation the mapping by composing nonlinear activation functions class integral...
The classical development of neural networks has primarily focused on learning mappings between finite dimensional Euclidean spaces or sets. We propose a generalization to learn operators, termed that map infinite function spaces. formulate the operator as composition linear integral operators and nonlinear activation functions. prove universal approximation theorem for our proposed operator, showing it can approximate any given continuous operator. are also discretization-invariant, i.e.,...
In this paper, we propose physics-informed neural operators (PINO) that combine training data and physics constraints to learn the solution operator of a given family parametric Partial Differential Equations (PDE). PINO is first hybrid approach incorporating PDE at different resolutions operator. Specifically, in PINO, coarse-resolution with imposed higher resolution. The resulting model can accurately approximate ground-truth for many popular families shows no degradation accuracy even...
In this article, we propose physics-informed neural operators (PINO) that combine training data and physics constraints to learn the solution operator of a given family parametric Partial Differential Equations (PDE). PINO is first hybrid approach incorporating PDE at different resolutions operator. Specifically, in PINO, coarse-resolution with imposed higher resolution. The resulting model can accurately approximate ground-truth for many popular families shows no degradation accuracy even...
Deep learning surrogate models have shown promise in solving partial differential equations (PDEs). Among them, the Fourier neural operator (FNO) achieves good accuracy, and is significantly faster compared to numerical solvers, on a variety of PDEs, such as fluid flows. However, FNO uses Fast transform (FFT), which limited rectangular domains with uniform grids. In this work, we propose new framework, viz., geo-FNO, solve PDEs arbitrary geometries. Geo-FNO learns deform input (physical)...
Nested FNO is a machine learning framework that offers general-purpose numerical simulator alternative to provide high-resolution CO 2 storage predictions in real time.
One of the main challenges in using deep learning-based methods for simulating physical systems and solving partial differential equations (PDEs) is formulating physics-based data desired structure neural networks. Graph networks (GNNs) have gained popularity this area since graphs offer a natural way modeling particle interactions provide clear discretizing continuum models. However, constructed approximating such tasks usually ignore long-range due to unfavorable scaling computational...
Vision transformers have delivered tremendous success in representation learning. This is primarily due to effective token mixing through self attention. However, this scales quadratically with the number of pixels, which becomes infeasible for high-resolution inputs. To cope challenge, we propose Adaptive Fourier Neural Operator (AFNO) as an efficient mixer that learns mix domain. AFNO based on a principled foundation operator learning allows us frame continuous global convolution without...
Bacteria can swim upstream in a narrow tube and pose clinical threat of urinary tract infection to patients implanted with catheters. Coatings structured surfaces have been proposed repel bacteria, but no such approach thoroughly addresses the contamination problem Here, on basis physical mechanism swimming, we propose novel geometric design, optimized by an artificial intelligence model. Using
Abstract Predicting plasma evolution within a Tokamak reactor is crucial to realizing the goal of sustainable fusion. Capabilities in forecasting spatio-temporal rapidly and accurately allow us quickly iterate over design control strategies on current devices future reactors. Modelling using numerical solvers often expensive, consuming many hours supercomputers, hence, we need alternative inexpensive surrogate models. We demonstrate accurate predictions both simulation experimental domains...
We propose the geometry-informed neural operator (GINO), a highly efficient approach to learning solution of large-scale partial differential equations with varying geometries. GINO uses signed distance function and point-cloud representations input shape operators based on graph Fourier architectures learn operator. The handles irregular grids transforms them into from regular latent which can be efficiently applied. is discretization-convergent, meaning trained model applied arbitrary...
Abstract Functional ultrasound imaging (fUSI) is a promising neuroimaging method that infers neural activity by detecting cerebral blood volume changes. It offers high sensitivity and spatial resolution relative to fMRI an epidural alternative electrophysiology for medical neuroscience applications, including brain-computer interfaces. However, current fUSI methods require hundreds of compounded images pulse emissions, leading computational costs, memory demands, potential probe heating. We...
Numerical simulation of multiphase flow in porous media is essential for many geoscience applications. Machine learning models trained with numerical data can provide a faster alternative to traditional simulators. Here we present U-FNO, novel neural network architecture solving problems superior accuracy, speed, and efficiency. U-FNO designed based on the newly proposed Fourier operator (FNO), which has shown excellent performance single-phase flows. We extend FNO-based highly complex...