A5100696195- Machine Learning and Data Classification
- Domain Adaptation and Few-Shot Learning
- Human Pose and Action Recognition
- Robotics and Sensor-Based Localization
- Multimodal Machine Learning Applications
- Anomaly Detection Techniques and Applications
- Hand Gesture Recognition Systems
- Gait Recognition and Analysis
- Quantum Information and Cryptography
- Cancer-related molecular mechanisms research
- Image and Object Detection Techniques
- Architecture and Computational Design
- Emotion and Mood Recognition
- Face and Expression Recognition
- Medical Image Segmentation Techniques
- Architecture, Design, and Social History
- Robotic Path Planning Algorithms
- Quantum Computing Algorithms and Architecture
- Natural Language Processing Techniques
- Click Chemistry and Applications
- Quantum Mechanics and Applications
- Educational Technology and Assessment
- Machine Learning and ELM
- Robot Manipulation and Learning
- Mycobacterium research and diagnosis
Northwest A&F University
2025
Institute of Software
2022-2025
Chinese Academy of Sciences
2023-2025
Tsinghua University
2025
Laboratoire d'Informatique, de Modélisation et d'Optimisation des Systèmes
2024
Université Clermont Auvergne
2024
University of Chinese Academy of Sciences
2022-2023
Shanghai Institute of Organic Chemistry
2023
Beijing University of Technology
2021-2022
Beijing University of Posts and Telecommunications
2010
Micro-expressions (MEs) are involuntary movements revealing people's hidden feelings, which has attracted numerous interests for its objectivity in emotion detection. However, despite wide applications various scenarios, micro-expression recognition (MER) remains a challenging problem real life due to three reasons, including (i) data-level: lack of data and imbalanced classes, (ii) feature-level: subtle, rapid changing, complex features MEs, (iii) decision-making-level: impact individual...
There is growing interest in covalent targeted inhibitors drug discovery against previously "undruggable" sites and targets. These molecules typically feature an electrophilic warhead that reacts with nucleophilic groups of protein residues, most notably the thiol group cysteines. One main challenge field to develop versatile utilizable warheads. Here, we characterize unique features novel arsenous warheads for reaction species a reversible manner further demonstrate organoarsenic probes can...
Fusarium head blight (FHB) is a devastating fungal disease in wheat, causing significant yield losses and deterioration of grain quality under severe conditions. In this study, genome-wide association study was conducted with 448 accessions using genotyping data generated by the 660K SNP array. Nine relatively stable FHB resistance loci were identified on chromosomes 1B, 1D, 2D, 5B, 7A 7B, respectively. Each QTL accounted for 4.1-10.4% phenotypic variation. Among them, QFhb.nwafu-7BS,...
Efficient recognition of emotions has attracted extensive research interest, which makes new applications in many fields possible, such as human-computer interaction, disease diagnosis, service robots, and so forth. Although existing work on sentiment analysis relying sensors or unimodal methods performs well for simple contexts like business recommendation facial expression recognition, it does far below expectations complex scenes, sarcasm, disdain, metaphors. In this article, we propose a...
In this paper, we explore the embedding of nonlinear dynamical systems into linear ordinary differential equations (ODEs) via Carleman linearization method. Under dissipative conditions, numerous previous works have established rigorous error bounds and convergence for linearization, which facilitated identification quantum advantages in simulating large-scale systems. Our analysis extends these findings by exploring beyond traditional condition, thereby broadening scope computational...
<p> Artificial intelligence technology has already had a profound impact in various fields such as economy, industry, and education, but still limited. Meta-learning, also known "learning to learn", provides an opportunity for general artificial intelligence, which can break through the current AI bottleneck. However, meta learning started late there are fewer projects compare with CV, NLP etc. Each deployment requires lot of experience configure environment, debug code or even...
Algorithms such as RRT (Rapidly exploring random tree), A* and their variants have been widely used in the field of robot path planning. A lot work has shown that these detectors are unable to carry out effective stable results for moving objects high-dimensional space, which generate a large number multi-dimensional corner points. Although some filtering mechanisms (such splines valuation functions) reduce calculation scale, chance collision is increased, fatal robots. In order fewer but...
Gestures, as a basic human feature, occupy an important position in human-computer interaction and other fields well. In order to accurately recognize gestures eliminate environmental interference, this paper proposes gesture recognition matching method based on dynamic bones. This uses the Mask R-CNN model exponential filtering identify calibrate key points of hand. Through segmentation feature extraction real-time frame images, combined network is used obtain stable accurate hand bone...
In recent years, self-supervised learning (SSL) has emerged as a promising approach for extracting valuable representations from unlabeled data. One successful SSL method is contrastive learning, which aims to bring positive examples closer while pushing negative apart. Many current approaches utilize parameterized projection head. Through combination of empirical analysis and theoretical investigation, we provide insights into the internal mechanisms head its relationship with phenomenon...
The goal of generality in machine learning is to achieve excellent performance on various unseen tasks and domains. Recently, self-supervised (SSL) has been regarded as an effective method this goal. It can learn high-quality representations from unlabeled data promising empirical multiple downstream tasks. Existing SSL methods mainly constrain two aspects: (i) large-scale training data, (ii) task-level shared knowledge. However, these lack explicit modeling the objective, theoretical...
Human Activity Recognition (HAR) is a field of study that focuses on identifying and classifying human activities. Skeleton-based has received much attention in recent years, where Graph Convolutional Network (GCN) based method widely used achieved remarkable results. However, the representation skeleton data issue over-smoothing GCN still need to be studied. 1). Compared central nodes, edge nodes can only aggregate limited neighbor information, different body are always structurally...
Meta-learning, or ”learning to learn”, enables machines acquire general priors with minimal supervision and rapidly adapt new tasks. Unlike traditional AI methods that approach each task from scratch using a fixed learning algorithm, meta-learning refines the algorithm itself through experience across various tasks, enhancing transferability generalization. This is especially valuable when data collection difficult costly, allowing for effective sequences while reducing dependency on...
An effective paradigm of multi-modal learning (MML) is to learn unified representations among modalities. From a causal perspective, constraining the consistency between different modalities can mine that convey primary events. However, such simple may face risk insufficient or unnecessary information: necessary but cause invariant across not have required accuracy; sufficient tends adapt well specific be hard new data. To address this issue, in paper, we aim are both and necessary, i.e.,...
Meta-learning has emerged as a powerful approach for leveraging knowledge from previous tasks to solve new tasks. The mainstream methods focus on training well-generalized model initialization, which is then adapted different with limited data and updates. However, it pushes the overfitting Previous mainly attributed this lack of used augmentations address issue, but they were by sufficient effective augmentation strategies. In work, we more fundamental ``learning learn'' strategy...
Scene recognition, particularly for aerial and underwater images, often suffers from various types of degradation, such as blurring or overexposure. Previous works that focus on convolutional neural networks have been shown to be able extract panoramic semantic features perform well scene recognition tasks. However, low-quality images still impede model performance due the inappropriate use high-level features. To address these challenges, we propose an adaptive selection mechanism identify...
Humans excel at adapting perceptions and actions to diverse environments, enabling efficient interaction with the external world. This adaptive capability relies on biological nervous system (BNS), which activates different brain regions for distinct tasks. Meta-learning similarly trains machines handle multiple tasks but a fixed network structure, not as flexible BNS. To investigate role of structure (FNS) in meta-learning, we conduct extensive empirical theoretical analyses, finding that...
Interior design is a complex and creative discipline involving aesthetics, functionality, ergonomics, materials science. Effective solutions must meet diverse requirements, typically producing multiple deliverables such as renderings drawings from various perspectives. Consequently, interior processes are often inefficient demand significant creativity. With advances in machine learning, generative models have emerged promising means of improving efficiency by creating designs text...