A5101596205- Advanced Graph Neural Networks
- Topic Modeling
- Adversarial Robustness in Machine Learning
- Natural Language Processing Techniques
- Domain Adaptation and Few-Shot Learning
- Carbon and Quantum Dots Applications
- Machine Learning and Algorithms
- Advanced Photocatalysis Techniques
- Web Data Mining and Analysis
- Speech Recognition and Synthesis
- Caching and Content Delivery
- Complex Network Analysis Techniques
- Data Quality and Management
- Advanced Text Analysis Techniques
- Speech and dialogue systems
- Nanoplatforms for cancer theranostics
Jiangxi College of Applied Technology
2024
University of California, Los Angeles
2022
Abstract Wearing face masks is the best way to stop spread of respiratory infections. However, if are not sterilized, changing them too frequently can actually increase risk cross‐contamination. Herein, construction an antipathogen photocatalytic mask with carbon vacancy‐modified nitride nanosheets (g‐C 3 N 4 ‐V C Ns) coated on non‐woven fabrics out layer mask, offering effective and long‐term protection against damaging pathogens when exposed light reported. The introduced vacancies found...
Adversarial attacks on graphs have attracted considerable research interests. Existing works assume the attacker is either (partly) aware of victim model, or able to send queries it. These assumptions are, however, unrealistic. To bridge gap between theoretical graph and real-world scenarios, in this work, we propose a novel more realistic setting: strict black-box attack, which has no knowledge about model at all not allowed any queries. design such an attack strategy, first generic filter...
This paper studies the problem of resolving data inconsistency from multiple sources in managing related to power equipment for China's state grid corporation. proposes automatically align inconsistent devices sources, i.e., same that have entries with different values each source, by HENGE, a HEtetrogeneous Network GEneration model. HENGE builds into heterogeneous graph, and captures complex physical semantic relationships among devices. combines both feature relational information improves...
Supervised fine-tuning (SFT) is a common method to enhance the tool calling capabilities of Large Language Models (LLMs), with training data often being synthesized. The current synthesis process generally involves sampling set tools, formulating requirement based on these and generating call statements. However, tools sampled randomly lack relevance, making them difficult combine thus reducing diversity data. Additionally, work overlooks coherence between turns dialogues, leading gap...
Unsupervised extractive summarization is an important technique in information extraction and retrieval. Compared with supervised method, it does not require high-quality human-labelled summaries for training thus can be easily applied documents different types, domains or languages. Most of existing unsupervised methods including TextRank PACSUM rely on graph-based ranking sentence centrality. However, this scorer directly end-to-end training, the positional-related prior assumption often...