A5100422091- Generative Adversarial Networks and Image Synthesis
- Advanced Image Processing Techniques
- Handwritten Text Recognition Techniques
- Advanced Graph Neural Networks
- Digital Media Forensic Detection
- Advanced Vision and Imaging
- Face and Expression Recognition
- Domain Adaptation and Few-Shot Learning
- Advanced Neural Network Applications
- Image Processing and 3D Reconstruction
- Neural Networks and Applications
- Advanced Image and Video Retrieval Techniques
- Image Retrieval and Classification Techniques
- Image and Signal Denoising Methods
- Computer Graphics and Visualization Techniques
- Bayesian Methods and Mixture Models
- Anomaly Detection Techniques and Applications
- Image Processing Techniques and Applications
- Music and Audio Processing
- Stock Market Forecasting Methods
- Text and Document Classification Technologies
- Adversarial Robustness in Machine Learning
- Explainable Artificial Intelligence (XAI)
- Machine Learning and Data Classification
- Multimodal Machine Learning Applications
Xi’an Jiaotong-Liverpool University
2015-2024
Changzhou University
2024
Zhejiang University of Science and Technology
2022
Shanghai Jiao Tong University
2022
Zhejiang University
2022
Tsinghua University
2021
The University of Melbourne
2019-2020
University of California, Irvine
2019
Chongqing University
2018
University of Chinese Academy of Sciences
2018
Stock index price prediction is prevalent in both academic and economic fields. The hard to forecast due its uncertain noise. With the development of computer science, neural networks are applied kinds industrial In this paper, we introduce four different methods machine learning including three typical models: Multilayer Perceptron (MLP), Long Short Term Memory (LSTM) Convolutional Neural Network (CNN) one attention-based network. main task predict next day’s according historical data....
Abstract We present a joint model based on deep learning that is designed to inpaint the missing-wedge sinogram of electron tomography and reduce residual artifacts in reconstructed tomograms. Traditional methods, such as weighted back projection (WBP) simultaneous algebraic reconstruction technique (SART), lack ability recover unacquired project information result limited tilt range; consequently, tomograms using these methods are distorted contaminated with elongation, streaking, ghost...
Graph convolutional networks (GCNs) emerge as the most successful learning models for graph-structured data. Despite their success, existing GCNs usually ignore entangled latent factors typically arising in real-world graphs, which results nonexplainable node representations. Even worse, while emphasis has been placed on local graph information, global knowledge of entire is lost to a certain extent. In this work, address these issues, we propose novel framework GCNs, termed LGD-GCN, taking...
We propose a continuous neural network architecture, termed Explainable Tensorized Neural Ordinary Differential Equations (ETN-ODE), for multi-step time series prediction at arbitrary points. Unlike the existing approaches, which mainly handle univariate or multivariate single-step prediction, ETN-ODE could model arbitrary-step prediction. In addition, it enjoys tandem attention, w.r.t. temporal attention and variable being able to provide explainable insights into data. Specifically,...
Spectral Graph Neural Networks (GNNs) have achieved tremendous success in graph machine learning, with polynomial filters applied for convolutions, where all nodes share the identical filter weights to mine their local contexts. Despite success, existing spectral GNNs usually fail deal complex networks (e.g., WWW) due such homogeneous filtering setting that ignores regional heterogeneity as typically seen real-world networks. To tackle this issue, we propose a novel diverse (DSF) framework,...
While exogenous variables have a major impact on performance improvement in time series analysis, interseries correlation and dependence among them are rarely considered the present continuous methods. The dynamical systems of multivariate could be modeled with complex unknown partial differential equations (PDEs) which play prominent role many disciplines science engineering. In this article, we propose continuous-time model for arbitrary-step prediction to learn an PDE system whose...
Generalised image outpainting is an important and active research topic in computer vision, which aims to extend appealing content all-side around a given image. Existing state-of-the-art methods often rely on discrete extrapolation the feature map bottleneck. They thus suffer from unsmoothness, especially circumstances where outlines of objects extrapolated regions are incoherent with input sub-images. To mitigate this issue, we design novel bottleneck Neural ODEs make continuous latent...
Style synthesis attracts great interests recently, while few works focus on its dual problem "style separation". In this paper, we propose the Separation and Synthesis Generative Adversarial Network (S3-GAN) to simultaneously implement style separation object photographs of specific categories. Based assumption that lie a manifold, contents styles are independent, employ S3-GAN build mappings between manifold latent vector space for separating synthesizing styles. The consists an encoder...
Diffusion models have demonstrated remarkable performance in image generation tasks, paving the way for powerful AIGC applications. However, these widely-used generative can also raise security and privacy concerns, such as copyright infringement, sensitive data leakage. To tackle issues, we propose a method, Unlearnable Perturbation, to safeguard images from unauthorized exploitation. Our approach involves designing an algorithm generate sample-wise perturbation noise each be protected....