A5103084324- Topic Modeling
- Adversarial Robustness in Machine Learning
- Anomaly Detection Techniques and Applications
- Generative Adversarial Networks and Image Synthesis
- Natural Language Processing Techniques
- Biomedical Text Mining and Ontologies
- Gaussian Processes and Bayesian Inference
- Information Retrieval and Search Behavior
- Advanced Graph Neural Networks
- Model Reduction and Neural Networks
- Image Retrieval and Classification Techniques
- Multimodal Machine Learning Applications
- Domain Adaptation and Few-Shot Learning
- Neural Networks and Applications
- Fault Detection and Control Systems
- Imbalanced Data Classification Techniques
- Machine Learning in Healthcare
- Digital Media Forensic Detection
- Data Quality and Management
- Advanced Malware Detection Techniques
- Text and Document Classification Technologies
- Markov Chains and Monte Carlo Methods
- Blind Source Separation Techniques
- Bayesian Methods and Mixture Models
- RNA modifications and cancer
Beijing Electronic Science and Technology Institute
2024
UNSW Sydney
2022-2024
Qufu Normal University
2013-2024
Kuaishou (China)
2024
China University of Mining and Technology
2024
Second Affiliated Hospital of Nanchang University
2023
Nanchang University
2023
University of Electronic Science and Technology of China
2023
China Medical University
2022
Jilin College of the Arts
2022
Adversarial perturbations of normal images are usually imperceptible to humans, but they can seriously confuse state-of-the-art machine learning models. What makes them so special in the eyes image classifiers? In this paper, we show empirically that adversarial examples mainly lie low probability regions training distribution, regardless attack types and targeted Using statistical hypothesis testing, find modern neural density models surprisingly good at detecting perturbations. Based on...
The imputation of missing values in time series has many applications healthcare and finance. While autoregressive models are natural candidates for imputation, score-based diffusion have recently outperformed existing counterparts including tasks such as image generation audio synthesis, would be promising imputation. In this paper, we propose Conditional Score-based Diffusion Imputation (CSDI), a novel method that utilizes conditioned on observed data. Unlike approaches, the conditional...
Adversarial examples are typically constructed by perturbing an existing data point within a small matrix norm, and current defense methods focused on guarding against this type of attack. In paper, we propose unrestricted adversarial examples, new threat model where the attackers not restricted to norm-bounded perturbations. Different from perturbation-based attacks, synthesize entirely scratch using conditional generative models. Specifically, first train Auxiliary Classifier Generative...
Energy-Based Models (EBMs), also known as non-normalized probabilistic models, specify probability density or mass functions up to an unknown normalizing constant. Unlike most other EBMs do not place a restriction on the tractability of constant, thus are more flexible parameterize and can model expressive family distributions. However, constant makes training particularly difficult. Our goal is provide friendly introduction modern approaches for EBM training. We start by explaining maximum...
Constructing Bi/BiOX (X = Cl, Br) heterostructures with unique electron transfer channels enables charge carriers to unidirectionally at the metal/semiconductor junction and inhibits backflow of photogenerated carriers. Herein, novel pine dendritic nanoassemblies multiple have been successfully synthesized assistance l-cysteine (l-Cys) through a one-step solvothermal method. Such Bi/BiOBr photocatalyst shows excellent activity toward degradation many antibiotics such as tetracycline (TC),...
Adversarial attacks often involve random perturbations of the inputs drawn from uniform or Gaussian distributions, e.g., to initialize optimization-based white-box generate update directions in black-box attacks. These simple perturbations, however, could be sub-optimal as they are agnostic model being attacked. To improve efficiency these attacks, we propose Output Diversified Sampling (ODS), a novel sampling strategy that attempts maximize diversity target model's outputs among generated...
In this paper, from the water spectral reflectance characteristics, I select five multi-spectral remote sensing information extraction methods, issue comparative analysis and applicability assessment in typical experiment areas based on Landsat data. have made a comprehensive comparison algorithm kinds of methods. The main conclusions are following: MNDWI relationship method between multi-band most widely applicability, always use building shadow bare land. has special effects mountain...
Learning generative models for graph-structured data is challenging because graphs are discrete, combinatorial, and the underlying distribution invariant to ordering of nodes. However, most existing not chosen ordering, which might lead an undesirable bias in learned distribution. To address this difficulty, we propose a permutation approach modeling graphs, using recent framework score-based modeling. In particular, design equivariant, multi-channel graph neural network model gradient at...
Abstract Background Long non-coding RNAs (lncRNAs) play an important role in angiogenesis, immune response, inflammatory response and tumor development metastasis. m6 A (N6—methyladenosine) is one of the most common RNA modifications eukaryotes. The aim our research was to investigate potential prognostic value m6A-related lncRNAs ovarian cancer (OC). Methods data we need for downloaded from Cancer Genome Atlas (TCGA) Gene Expression Omnibus (GEO) database. Pearson correlation analysis...
We propose a new family of efficient and expressive deep generative models graphs, called Graph Recurrent Attention Networks (GRANs). Our model generates graphs one block nodes associated edges at time. The size sampling stride allow us to trade off sample quality for efficiency. Compared previous RNN-based graph models, our framework better captures the auto-regressive conditioning between already-generated to-be-generated parts using Neural (GNNs) with attention. This not only reduces...
Several machine learning applications involve the optimization of higher-order derivatives (e.g., gradients gradients) during training, which can be expensive in respect to memory and computation even with automatic differentiation. As a typical example generative modeling, score matching (SM) involves trace Hessian. To improve computing efficiency, we rewrite SM objective its variants terms directional derivatives, present generic strategy efficiently approximate any-order derivative finite...