A5102748847- Matrix Theory and Algorithms
- Quantum Computing Algorithms and Architecture
- Sparse and Compressive Sensing Techniques
- Quantum Information and Cryptography
- Electromagnetic Scattering and Analysis
- Advanced Chemical Physics Studies
- Neural Networks and Applications
- Machine Learning in Materials Science
- Model Reduction and Neural Networks
- Stochastic Gradient Optimization Techniques
- Tensor decomposition and applications
- Quantum and electron transport phenomena
- Parallel Computing and Optimization Techniques
- Advanced Optimization Algorithms Research
- Spectroscopy and Quantum Chemical Studies
- Advanced Image Processing Techniques
- Generative Adversarial Networks and Image Synthesis
- Image and Signal Denoising Methods
- Mathematical Approximation and Integration
- Advanced NMR Techniques and Applications
- Neural Networks and Reservoir Computing
- Numerical methods for differential equations
- Iterative Methods for Nonlinear Equations
- Advanced Neural Network Applications
- Machine Learning and ELM
Fudan University
2017-2025
Shaoxing People's Hospital
2025
Qilu University of Technology
2020-2024
Shandong Academy of Sciences
2020-2024
Applied Mathematics (United States)
2024
Duke University
2018-2020
Stanford University
2015-2017
Strong catalyst–support interaction plays a key role in heterogeneous catalysis, as has been well-documented high-temperature gas-phase chemistry, such the water gas shift reaction. Insight into how interactions can be exploited to optimize catalytic activity aqueous electrochemistry, however, is still lacking. In this work, we show rationally designed electrocatalyst/support interface greatly impact overall electrocatalytic of Ni–Fe layered double hydroxide (NiFeLDH) oxidation. particular,...
The paper introduces the butterfly factorization as a data-sparse approximation for matrices that satisfy complementary low-rank property. can be constructed efficiently if either fast algorithms applying matrix and its adjoint are available or entries of sampled individually. For an $N\times N$ matrix, resulting is product $O(\log N)$ sparse matrices, each with $O(N)$ nonzero entries. Hence, it applied rapidly in $O(N\log operations. Numerical results provided to demonstrate effectiveness...
Near-term quantum computers will be limited in the number of qubits on which they can process information as well depth circuits that coherently carry out. To date, experimental demonstrations algorithms such Variational Quantum Eigensolver (VQE) have been to small molecules using minimal basis sets for this reason. In work we propose incorporating an orbital optimization scheme into eigensolvers wherein a parametrized partial unitary transformation is applied functions set order reduce...
We develop an efficient algorithm, coordinate descent FCI (CDFCI), for the electronic structure ground-state calculation in configuration interaction framework. CDFCI solves unconstrained nonconvex optimization problem, which is a reformulation of full eigenvalue via adaptive method with deterministic compression strategy. captures and updates appreciative determinants different frequencies proportional to their importance. show that produces accurate variational energy both static dynamic...
First-principles electronic structure calculations are very widely used thanks to the many successful software packages available. Their traditional coding paradigm is monolithic, i.e., regardless of how modular its internal may be, code built independently from others, compiler up, with exception linear-algebra and message-passing libraries. This model has been quite for decades. The rapid progress in methodology, however, resulted an ever increasing complexity those programs, which implies...
This paper introduces a novel approach to implementing non-unitary linear transformations of basis on quantum computational platforms, significant leap beyond the conventional unitary methods. By integrating Singular Value Decomposition (SVD) into process, method achieves an operational depth $O(n)$ with about $n$ ancilla qubits, enhancing capabilities for analyzing fermionic systems. The non-unitarity transformation allows us transform wave function from one another, which can span...
We develop a multithreaded parallel coordinate descent full configuration interaction algorithm (mCDFCI) for the electronic structure ground-state calculation in framework. The FCI problem is reformulated as an unconstrained minimization and tackled by modified block method with deterministic compression strategy. mCDFCI designed to prioritize determinants based on their importance, updates enabling efficient parallelization shared-memory, multicore computing infrastructure. demonstrate...
To compare the bone cement diffusion and clinical effects between conventional percutaneous vertebroplasty(PVP) application of positioning reduction targeted puncture techniques in treatment elderly patients with osteoporotic vertebral compression fractures. A retrospective comparative study was conducted, analyzing data 268 single-level fractures admitted January 2021 March 2023. The were divided into two groups:the PVP group (138 cases) (130 cases). Among them, 138 treated by traditional...
We present an efficient way to compute the excitation energies in molecules and solids within linear-response time-dependent density functional theory (LR-TDDFT). Conventional methods construct diagonalize LR-TDDFT Hamiltonian require ultrahigh computational cost, limiting its optoelectronic applications small systems. Our new method is based on interpolative separable fitting (ISDF) decomposition combined with implicitly constructing iteratively diagonalizing only requires low cost...
We propose in this work RBM-SVGD, a stochastic version of Stein Variational Gradient Descent (SVGD) method for efficiently sampling from given probability measure and thus useful Bayesian inference. The is to apply the Random Batch Method (RBM) interacting particle systems proposed by Jin et al SVGD. While keeping behaviors SVGD, it reduces computational cost, especially when kernel has long range. Numerical examples verify efficiency new
Leading eigenvalue problems for large scale matrices arise in many applications. Coordinatewise descent methods are considered this work such based on a reformulation of the leading problem as nonconvex optimization problem. The convergence several coordinatewise is analyzed and compared. Numerical examples applications to quantum many-body demonstrate efficiency provide benchmarks proposed methods.
We consider universal approximations of symmetric and anti-symmetric functions, which are important for applications in quantum physics, as well other scientific engineering computations. give constructive with explicit bounds on the number parameters respect to dimension target accuracy $\epsilon$. While approximation still suffers from curse dimensionality, best our knowledge, these first results literature error functions symmetry or anti-symmetry constraints.
The hierarchical interpolative factorization (HIF) offers an efficient way for solving or preconditioning elliptic partial differential equations. By exploiting locality and low-rank properties of the operators, HIF achieves quasi-linear complexity factorizing discrete positive definite operator linear associated system. In this paper, distributed-memory (DHIF) is introduced as a parallel implementation HIF. DHIF organizes processes in structure keep communication local possible. computation...
This paper introduces the interpolative butterfly factorization for nearly optimal implementation of several transforms in harmonic analysis, when their explicit formulas satisfy certain analytic properties and matrix representations these a complementary low-rank property. A preliminary is constructed based on approximations matrix. novel sweeping compression technique further compresses via sequence structure-preserving approximations. The procedure propagates property among neighboring...
The minority carrier diffusion length (LD) is a crucial property that determines the performance of light absorbers in photoelectrochemical (PEC) cells. Many transition-metal oxides are stable photoanodes for solar water splitting but exhibit small to moderate LD, ranging from few nanometers (such as α-Fe2O3 and TiO2) tens BiVO4). Under operating conditions, temperature PEC cells can deviate substantially ambient, yet dependence LD has not been quantified. In this work, we show measuring...
Kernel methods are widespread in machine learning; however, they limited by the quadratic complexity of construction, application, and storage kernel matrices. Low-rank matrix approximation algorithms widely used to address this problem reduce arithmetic cost. However, we observed that for some datasets with wide intraclass variability, optimal parameter smaller classes yields a is less well-approximated low-rank methods. In paper, propose an efficient structured method---the block basis...
Low-rank approximations are popular methods to reduce the high computational cost of algorithms involving large-scale kernel matrices. The success low-rank hinges on matrix rank matrix, and in practice, these effective even for high-dimensional datasets. Their practical motivates our analysis function rank, an upper bound rank. In this paper, we consider radial basis functions (RBF), approximate RBF with a representation that is finite sum separate products, provide explicit bounds...
Deep networks, especially convolutional neural networks (CNNs), have been successfully applied in various areas of machine learning as well to challenging problems other scientific and engineering fields. This paper introduces Butterfly-Net, a low-complexity CNN with structured sparse cross-channel connections, together Butterfly initialization strategy for family networks. Theoretical analysis the approximation power Butterfly-Net Fourier representation input data shows that error decays...