A5100453848- Domain Adaptation and Few-Shot Learning
- Geometric Analysis and Curvature Flows
- Advanced Neural Network Applications
- Coding theory and cryptography
- Fluid Dynamics and Turbulent Flows
- Complex Network Analysis Techniques
- Multimodal Machine Learning Applications
- COVID-19 diagnosis using AI
- Topic Modeling
- Advanced Computational Techniques and Applications
- Cryptography and Data Security
- Natural Language Processing Techniques
- Anomaly Detection Techniques and Applications
- Advanced Graph Neural Networks
- Mobile Ad Hoc Networks
- Imbalanced Data Classification Techniques
- Stochastic processes and financial applications
- Network Security and Intrusion Detection
- Bioinformatics and Genomic Networks
- Geochemistry and Geologic Mapping
- Machine Learning and ELM
- Blockchain Technology Applications and Security
- graph theory and CDMA systems
- Advanced Neuroimaging Techniques and Applications
- Error Correcting Code Techniques
Southern Medical University
2025
Middle Tennessee State University
2021-2024
The University of Tokyo
2011-2024
Georgia Institute of Technology
2022-2023
University of California, Los Angeles
2023
Central South University
2022
Beijing University of Posts and Telecommunications
2021-2022
Hong Kong Polytechnic University
2022
Shanghai Electric (China)
2020-2021
Southeast University
2020
Exogenous reactive oxygen species (ROS) produced by nanobubble (NB) water offer a reasonable explanation for NBs' physiological promotion and oxidation effects. To develop exploit the NB technology, we have performed further research to identify specific ROS NBs. Using fluorescent reagent APF, Fenton reaction, dismutation reaction of superoxide dismutase DMSO, distinguished four types (superoxide anion radical (O2·-), hydrogen peroxide (H2O2), hydroxyl (·OH), singlet (1O2)). ·OH was...
Aspect sentiment triplet extraction (ASTE) is a challenging subtask in aspect-based analysis. It aims to explore the triplets of aspects, opinions and sentiments with complex correspondence from context. The bidirectional machine reading comprehension (BMRC), can effectively deal ASTE task, but several problems remains, such as query conflict probability unilateral decrease. Therefore, this paper presents robustly optimized BMRC method by incorporating four improvements. word segmentation...
ABSTRACT Objective This study aimed to identify genetic variants and their functional consequences underlying Unexplained Recurrent Pregnancy Loss (uRPL) through comprehensive genomic transcriptomic analyses. Methods We recruited 13 Chinese uRPL patients performed Whole Exome Sequencing (WES) on chorionic villi samples from miscarriage tissues. Additionally, we conducted an integrative analysis using single‐cell RNA sequencing data decidual immune cells examine expression patterns. Results...
.We study the Wasserstein Hamiltonian flow with a common noise on density manifold of finite graph. Under framework stochastic variational principle, we first develop formulation and show local existence unique solution. We also establish sufficient condition for global Consequently, obtain well-posedness nonlinear Schrödinger equations In addition, using Wong–Zakai approximation noise, prove minimizer an optimal control problem noise. that its satisfies graph as well.Keywordsstochastic...
In this paper, we develop and analyze numerical methods for high-dimensional Fokker--Planck equations by leveraging generative models from deep learning. Our starting point is a formulation of the equation as system ordinary differential (ODEs) on finite-dimensional parameter space with parameters inherited such normalizing flows. We call ODEs neural parametric equations. The fact that can be viewed $L^2$-Wasserstein gradient flow Kullback--Leibler (KL) divergence allows us to derive...
Deep learning algorithms face great challenges with long-tailed data distribution which, however, is quite a common case in real-world scenarios. Previous methods tackle the problem from either aspect of input space (re-sampling classes different frequencies) or loss (re-weighting weights), suffering heavy over-fitting to tail hard optimization during training. To alleviate these issues, we propose more fundamental perspective for recognition, i.e., parameter space, and aims preserve...
Previous adversarial training raises model robustness under the compromise of accuracy on natural data. In this paper, we reduce degradation. We use logits from one clean to guide learning another robust model, taking into consideration that well trained embed most discriminative features data, {\it e.g.}, generalizable classifier boundary. Our solution is constrain takes examples as input and makes it similar those fed with corresponding It lets inherit boundary model. Moreover, observe...
Efficient list decoding of rank-metric codes seems more difficult compared with classical block although decodability random is completely determined by Ding. For example, it was shown Raviv and Wachter-Zeh that the radius Gabidulin same as unique radius, i.e., half minimum distance for some instances parameters. On other hand, Guruswami Xing give an explicit construction subcodes codes, which can be decoded up to Singleton bound. This implies are good candidates decoding. In this paper, we...
Federated learning is a framework for machine that dedicated to data privacy protection. In federated learning, system cannot fully control the behavior of clients which can be faulty. These behaviors include sharing arbitrary faulty gradients and delaying process due Byzantine attacks or clients' own software hardware failures. parameter server may also during gradient collection aggregation, mainly including gradient-based training inference model update. The above problems lead reduced...
Training semantic segmentation models requires a large amount of finely annotated data, making it hard to quickly adapt novel classes not satisfying this condition. Few-Shot Segmentation (FS-Seg) tackles problem with many constraints. In paper, we introduce new benchmark, called Generalized Semantic (GFS-Seg), analyze the generalization ability simultaneously segmenting categories very few examples and base sufficient examples. It is first study showing that previous representative...
In foreign exchange (FX) markets, the key issues to achieve profitable trading rules are combination of indicators, selection their parameters, and decision trade timing for orders settlements. this paper, we present a system using genetic algorithm (GA) programming (GP). Unlike related researches on problem, our work focuses two aspects. First, calculation appropriate settlement is proposed, make more profits less losses. Second, reverse trend data generated in-sample data, overcome...
PID and fuzzy controller are used for the wind turbine with nonlinear mathematical model. control has simple principle its parameters setting rather easy. Fuzzy combines reasoning goals/constraints on robustness (robust controller), which can effectively improve robustness. MATLAB software is simulations, results display that robust better performances than when errors exist, especially performance.
Abstract Motivation Understanding the dynamics of gene regulatory networks (GRNs) across various cellular states is crucial for deciphering underlying mechanisms governing cell behavior and functionality. However, current comparative analytical methods, which often focus on simple topological information such as degree genes, are limited in their ability to fully capture similarities differences among complex GRNs. Results We present Gene2role, a embedding approach that leverages multi-hop...
The musculoskeletal network is a complex system of different types nodes and edges interacting with each other. Although there wealth knowledge about the anatomical components human body connections between them, interdependence these as remains largely unexplored. This study aims to understand structure networks by using hypergraphs model many-to-many connections. We used both pairwise hypergraph-based embedding methods learn connectivity muscles. Experiments demonstrated superiority...
Cloud application provides access to large pool of virtual machines for building high-quality applications satisfy customers’ requirements. A difficult issue is how predict machine response time because it determines when we could adjust dynamic scalable machines. To address the critical issue, this paper proposes a prediction method which based on genetic algorithm-back propagation (GA-BP) neural network. First all, component by past usage experience data: number concurrent requests and...