A5111256957- Advanced Graph Neural Networks
- Topic Modeling
- Seismic Imaging and Inversion Techniques
- AI in cancer detection
- Domain Adaptation and Few-Shot Learning
- Radiomics and Machine Learning in Medical Imaging
- Advanced Steganography and Watermarking Techniques
- Regional Development and Environment
- Speech and Audio Processing
- Multi-Criteria Decision Making
- Electrohydrodynamics and Fluid Dynamics
- Advanced SAR Imaging Techniques
- Text and Document Classification Technologies
- Infrared Target Detection Methodologies
- BIM and Construction Integration
- Optimal Experimental Design Methods
- Adversarial Robustness in Machine Learning
- Advanced Optical Sensing Technologies
- Robot Manipulation and Learning
- Advanced Multi-Objective Optimization Algorithms
- Icing and De-icing Technologies
- Radar Systems and Signal Processing
- Design Education and Practice
- Energy, Environment, Economic Growth
- Advanced Neural Network Applications
Cedars-Sinai Medical Center
2024-2025
Hong Kong Polytechnic University
2024
University of Hong Kong
2024
George Washington University
2024
Osaka University
2024
Shanghai First People's Hospital
2024
Shanghai Jiao Tong University
2024
Shanghai Electric (China)
2024
Ocean University of China
2024
National University of Defense Technology
2023-2024
We introduce SONO, a novel method leveraging Second-Order Neural Ordinary Differential Equations (Second-Order NODEs) to enhance cross-modal few-shot learning. By employing simple yet effective architecture consisting of NODEs model paired with classifier, SONO addresses the significant challenge overfitting, which is common in scenarios due limited training examples. Our second-order approach can approximate broader class functions, enhancing model's expressive power and feature...
Fabrication technologies based on electro-hydrodynamic processes have been extensively studied in the past decades. Near-field electrospinning (NFES), a stable cone-jet mode, is widely used to fabricate micro- and nano-scale fibrous structures for variety of applications. However, previous reviews given limited attention capabilities NFES 2D 3D structures. This review introduces four key metrics capabilities, i.e., fidelity, resolution, response, aspect ratio, evaluate summarize advances...
The additive model is a popular nonparametric regression method due to its ability retain modeling flexibility while avoiding the curse of dimensionality. backfitting algorithm an intuitive and widely used numerical approach for fitting models. However, application large datasets may incur high computational cost thus infeasible in practice. To address this problem, we propose novel called independence-encouraging subsampling (IES) select subsample from big data training Inspired by minimax...
Depression can be reflected by long-term human spatio-temporal facial behaviours. While face videos recorded in real-world usually have long and variable lengths, existing video-based depression assessment approaches frequently re-sample/down-sample such to short equal-length videos, or split each video into several segments, where segment-level behaviours are suppressed as a vector-style representations for RNN-based (video-level) modelling. Both strategies lead crucial information loss...
Deceptive path planning (DPP) aims to find a that minimizes the probability of observer identifying real goal observed before it reaches. It is important for addressing issues such as public safety, strategic planning, and logistics route privacy protection. Existing traditional methods often rely on “dissimulation”—hiding truth—to obscure paths while ignoring time constraints. Building upon theory probabilistic recognition based cost difference, we proposed DPP method, DPP_Q, count-based...
Abstract Background Acute respiratory distress syndrome (ARDS) after cardiac surgery is a severe complication with high mortality and morbidity. Traditional clinical approaches may lead to under recognition of this heterogeneous syndrome, potentially resulting in diagnosis delay. This study aims develop external validate seven machine learning (ML) models, trained on electronic health records data, for predicting ARDS surgery. Methods multicenter, observational cohort included patients who...
Crude oil plays a significant role in the modern society and its price prediction attracts more attentions, not only for importance to industry, but also complex movement. Based on PMRS, ECM NN, this paper presents an integrated model forecast crude prices. In proposed model, PMRS is first used trend of price, then offered establish forecasting errors. Finally, NN employed integrate results from ones make final values accurate desirable. The WTI spot prices set financial indicators are...
Knowledge graphs (KGs) serve as useful resources for various applications including machine learning, data mining, and artificial intelligence. Graph Completion (KGC) aims at reasoning over known (observed) facts infer the missing (unobserved) relationships, to improve KG's coverage. Many recent works have laid stress on leveraging Convolutional Networks (GCNs) KGC task, which obtain good representation learning ability graph-structured data, achieved satisfactory results. However, GCN-based...
Large language models (LLMs) have garnered sub-stantial attention and significantly transformed the landscape of artificial intelligence, due to their human-like understanding generation capabilities. However, despite excellent capabilities, LLMs lack latest information are constrained by limited context memory, which limits effectiveness in many real-time applications that require up-to-date information, such as personal AI assistants. Inspired recent study on enhancing with infinite...
We introduce Motion-I2V, a novel framework for consistent and controllable image-to-video generation (I2V). In contrast to previous methods that directly learn the complicated mapping, Motion-I2V factorizes I2V into two stages with explicit motion modeling. For first stage, we propose diffusion-based field predictor, which focuses on deducing trajectories of reference image's pixels. second motion-augmented temporal attention enhance limited 1-D in video latent diffusion models. This module...
Heterogeneous Graphs (HGs) can effectively model complex relationships in the real world by multi-type nodes and edges. In recent years, inspired self-supervised learning, contrastive Neural Networks (HGNNs) have shown great potential utilizing data augmentation discriminators for downstream tasks. However, is still limited due to discrete abstract nature of graphs. To tackle above limitations, we propose a novel \textit{Generative-Contrastive Graph Network (GC-HGNN)}. Specifically, first...
In this paper, we consider the problem of prototype-based vision-language reasoning problem. We observe that existing methods encounter three major challenges: 1) escalating resource demands and prolonging training times, 2) contending with excessive learnable parameters, 3) fine-tuning based only on a single modality. These challenges will hinder their capability to adapt Vision-Language Models (VLMs) downstream tasks. Motivated by critical observation, propose novel method called...
Self-supervised learning (SSL) has recently attracted significant attention in the field of recommender systems. Contrastive (CL) stands out as a major SSL paradigm due to its robust ability generate self-supervised signals. Mainstream graph contrastive (GCL)-based methods typically implement CL by creating views through various data augmentation techniques. Despite these are effective, we argue that there still exist several challenges: i) Data (e.g., discarding edges or adding noise)...
We introduce SONO, a novel method leveraging Second-Order Neural Ordinary Differential Equations (Second-Order NODEs) to enhance cross-modal few-shot learning. By employing simple yet effective architecture consisting of NODEs model paired with classifier, SONO addresses the significant challenge overfitting, which is common in scenarios due limited training examples. Our second-order approach can approximate broader class functions, enhancing model's expressive power and feature...