A5103110744- Machine Learning and Data Classification
- Electrocatalysts for Energy Conversion
- Domain Adaptation and Few-Shot Learning
- Data Stream Mining Techniques
- Machine Learning and Algorithms
- Ammonia Synthesis and Nitrogen Reduction
- Catalytic Processes in Materials Science
- Face and Expression Recognition
- Advanced Graph Neural Networks
- Advanced Photocatalysis Techniques
- Adversarial Robustness in Machine Learning
- Algorithms and Data Compression
- Explainable Artificial Intelligence (XAI)
- Anomaly Detection Techniques and Applications
- Advanced Image and Video Retrieval Techniques
- Graph Theory and Algorithms
- Catalysis and Oxidation Reactions
- Text and Document Classification Technologies
- CO2 Reduction Techniques and Catalysts
- Advanced Neural Network Applications
- Insurance, Mortality, Demography, Risk Management
- Video Surveillance and Tracking Methods
- Ionic liquids properties and applications
- Copper-based nanomaterials and applications
- Image Retrieval and Classification Techniques
Beijing Advanced Sciences and Innovation Center
2025
Beijing University of Chemical Technology
2024-2025
Hefei University of Technology
2015-2022
Many graph-based semi-supervised learning methods for large datasets have been proposed to cope with the rapidly increasing size of data, such as Anchor Graph Regularization (AGR). This model builds a regularization framework by exploring underlying structure whole dataset both datapoints and anchors. Nevertheless, AGR still has limitations in its two components: (1) anchor graph construction, estimation local weights between each datapoint neighboring anchors could be biased relatively...
Several models have been proposed to cope with the rapidly increasing size of data, such as Anchor Graph Regularization (AGR). The AGR approach significantly accelerates graph-based learning by exploring a set anchors. However, when dataset becomes much larger, still faces big graph which brings dramatically computational costs. To overcome this issue, we propose novel Hierarchical (HAGR) multiple-layer anchors pyramid-style structure. In HAGR, labels datapoints are inferred from coarsest...
Machine learning can provide deep insights into data, allowing machines to make high-quality predictions and having been widely used in real-world applications, such as text mining, visual classification, recommender systems. However, most sophisticated machine approaches suffer from huge time costs when operating on large-scale data. This issue calls for the need of Large-scale Learning (LML), which aims learn patterns big data with comparable performance efficiently. In this paper, we...
We study the problem of active learning for multi-class classification on large-scale datasets. In this setting, existing approaches built upon uncertainty measures are ineffective discovering unknown regions, and those based expected error reduction inefficient owing to their huge time costs. To overcome above issues, paper proposes a novel query selection criterion called approximated (AER). AER, each candidate is estimated an impact over all datapoints ratio between its nearby datapoints....
Improving the electrocatalytic conversion of formate in alkaline solutions is crucial for commercial application fuel cells. However, palladium-based catalysts used oxidation reactions (FOR) face challenges due to strong adsorption hydrogen intermediates, resulting lower catalytic efficiency environments. Herein, we prepared a PdZr/C catalyst aimed at employing doping-induced strain strategy reduce binding energy palladium and release more active sites formate. Through density functional...
The electrochemical nitrogen oxidation reaction (NOR) presents a sustainable pathway for nitrate synthesis under mild conditions; however, the process is hindered by inadequate adsorption and activation of N2 on electrocatalysts. In this study, we utilized Co3O4 as model catalyst engineered lattice distortions introducing oxygen vacancies, which expanded eg band active sites to enhance activation. modified achieved Faradaic efficiency 10.68% yield 58.80 μg·h-1·mgcat-1. Comprehensive...
Over the last decade, deep neural networks (DNNs) are regarded as black-box methods, and their decisions criticized for lack of explainability. Existing attempts based on local explanations offer each input a visual saliency map, where supporting features that contribute to decision emphasized with high relevance scores. In this paper, we improve map differentiated explanations, which not only distinguishes from backgrounds but also shows different degrees importance various parts within...
Knowledge graphs have received intensive research interests. When the labels of most nodes or datapoints are missing, anchor graph and hierarchical models can be employed. With an graph, we only need to optimize coarsest anchors, inferred from these anchors in a coarse-to-fine manner. The complexity optimization is therefore reduced cubic cost with respect number anchors. However, obtain high accuracy when data distribution complex, scale this set still needs large, which thus inevitably...
Recent studies indicate that deep neural networks (DNNs) are prone to show discrimination towards certain demographic groups. We observe algorithmic can be explained by the high reliance of models on fairness sensitive features. Motivated this observation, we propose achieve suppressing DNN from capturing spurious correlation between those features with underlying task. Specifically, firstly train a bias-only teacher model which is explicitly encouraged maximally employ for prediction. The...
Curriculum Learning (CL) selects the training samples from easy to difficult boost classification results. Most existing variates of CL measure difficulty level an example in intuitive way, i.e., loss value between prediction and ground truth. This way ignores different distances every class boundary, which is implied vectors. In this paper, we propose a novel framework, named Balance Loss Learning(BLCL), reveal comprehensive improve curriculum process based on deep architectures. We follow...
The electrochemical synthesis of nitrates through nitrogen oxidation reactions (NOR) faces challenges due to the chemical stability N2 and complexity its electron transfer mechanisms. Here, we report enhancement NOR via Al doping into Co3O4. This modifies electronic structure Co3O4, particularly by increasing eg orbital occupation, which effectively activates N2. Our experimental results, corroborated density functional theory (DFT) calculations, reveal that shortens Co–N bond lengths...
Recent research has shown Deep Neural Networks (DNNs) to be vulnerable adversarial examples that induce desired misclassifications in the models. Such risks impede application of machine learning security-sensitive domains. Several defense methods have been proposed against attacks detect at test time or make models more robust. However, while existing are quite effective under blackbox threat model, where attacker is not aware defense, they relatively ineffective whitebox full knowledge...
Machine learning can provide deep insights into data, allowing machines to make high-quality predictions and having been widely used in real-world applications, such as text mining, visual classification, recommender systems. However, most sophisticated machine approaches suffer from huge time costs when operating on large-scale data. This issue calls for the need of {Large-scale Learning} (LML), which aims learn patterns big data with comparable performance efficiently. In this paper, we...