A5106732295- Advanced Graph Neural Networks
- Imbalanced Data Classification Techniques
- Topic Modeling
- Natural Language Processing Techniques
- Text and Document Classification Technologies
- Electricity Theft Detection Techniques
- Asymmetric Synthesis and Catalysis
- Explainable Artificial Intelligence (XAI)
- Information Systems and Technology Applications
- Advanced Text Analysis Techniques
- Graph Theory and Algorithms
- Blind Source Separation Techniques
- Neural Networks and Applications
- Spam and Phishing Detection
- Brain Tumor Detection and Classification
- Adversarial Robustness in Machine Learning
- Software Engineering Research
- Fluorine in Organic Chemistry
- Fractal and DNA sequence analysis
- Organic and Inorganic Chemical Reactions
University of Toronto
2022-2024
Fudan University
2023
Along with the rapid evolution of mobile communication technologies, such as 5G, there has been a significant increase in telecom fraud, which severely dissipates individual fortune and social wealth. In recent years, graph mining techniques are gradually becoming mainstream solution for detecting fraud. However, imbalance problem, caused by Pareto principle, brings severe challenges to data mining. This emerging complex issue received limited attention prior research. this paper, we propose...
In recent years, the increasing prevalence of mobile social network fraud has led to significant distress and depletion personal wealth, resulting in considerable economic harm. Graph neural networks (GNNs) have emerged as a popular approach tackle this issue. However, challenge graph imbalance, which can greatly impede effectiveness GNN-based detection methods, received little attention prior research. Thus, we are going present novel cost-sensitive (CSGNN) article. Initially, reinforcement...
The unique biological activity of pyrrolidine compounds has attracted significant interest from medicinal chemists. This study presents a one-step synthesis [indolyl-3,2'-pyrrolidine] with three chiral centers and an α-CF3 group in excellent yields enantioselectivities by reacting trifluoroethyl ketoimine unsaturated alkynyl ketone through guanidine catalized [3+2] cycloaddition reaction. mild reaction condition, low catalyst loading high rate make this synthetic method highly attractive for...
With the rapid evolution of mobile communication networks, number subscribers and their practices is increasing dramatically worldwide. However, fraudsters are also sniffing out benefits. Detecting from massive volume call detail records (CDR) in networks has become an important yet challenging topic. Fortunately, Graph neural network (GNN) brings new possibilities for telecom fraud detection. presence graph imbalance GNN oversmoothing problems makes fraudster detection unsatisfactory. To...
Document-level relation extraction (DocRE) attracts more research interest recently. While models achieve consistent performance gains in DocRE, their underlying decision rules are still understudied: Do they make the right predictions according to rationales? In this paper, we take first step toward answering question and then introduce a new perspective on comprehensively evaluating model.Specifically, conduct annotations provide rationales considered by humans DocRE. Then, investigations...
In recent years, an increasing number of telecom frauds have caused huge losses to people around the world. Graph neural network(GNN) brings new possibilities for fraud detection. However, due existence imbalance problem in graph, it is difficult general GNN models detect a small positive samples. To address this problem, we design GNN-based detector. First, transform node features with multilayer perceptron. Subsequently, reinforcement learning-based neighbor sampling strategy designed...
We propose yet another entity linking model (YELM) which links words to entities instead of spans. This overcomes any difficulties associated with the selection good candidate mention spans and makes joint training detection (MD) disambiguation (ED) easily possible. Our is based on BERT produces contextualized word embeddings are trained against a MD ED objective. achieve state-of-the-art results several standard (EL) datasets.
With the rapid development of mobile networks, people's social contacts have been considerably facilitated. However, rise network fraud upon those has caused a great deal distress, in case depleting personal and wealth, then potentially doing significant economic harm. To detect fraudulent users, call detail record (CDR) data, which portrays behavior users widely utilized. But imbalance problem aforementioned could severely hinder effectiveness detectors based on graph neural networks(GNN),...
Along with the rapid evolution of mobile communication technologies, such as 5G, there has been a drastically increase in telecom fraud, which significantly dissipates individual fortune and social wealth. In recent years, graph mining techniques are gradually becoming mainstream solution for detecting fraud. However, imbalance problem, caused by Pareto principle, brings severe challenges to data mining. This is new challenging but little previous work noticed. this paper, we propose Graph...
Document-level relation extraction (DocRE) attracts more research interest recently. While models achieve consistent performance gains in DocRE, their underlying decision rules are still understudied: Do they make the right predictions according to rationales? In this paper, we take first step toward answering question and then introduce a new perspective on comprehensively evaluating model. Specifically, conduct annotations provide rationales considered by humans DocRE. Then, investigations...
Gender bias in language models has attracted sufficient attention because it threatens social justice. However, most of the current debiasing methods degraded model's performance on other tasks while degradation mechanism is still mysterious. We propose a theoretical framework explaining three candidate mechanisms gender bias. use our to explain why cause degradation. also discover pathway through which will not degrade model performance. further develop causality-detection fine-tuning...