A5100734672- Topic Modeling
- Domain Adaptation and Few-Shot Learning
- Recommender Systems and Techniques
- Advanced Graph Neural Networks
- Face and Expression Recognition
- Machine Learning and ELM
- Text and Document Classification Technologies
- Data Mining Algorithms and Applications
- Image Retrieval and Classification Techniques
- Imbalanced Data Classification Techniques
- Natural Language Processing Techniques
- Machine Learning and Data Classification
- Multimodal Machine Learning Applications
- Speech Recognition and Synthesis
- Data Management and Algorithms
- Scientific Measurement and Uncertainty Evaluation
- Advanced Image and Video Retrieval Techniques
- Advanced Bandit Algorithms Research
- Microbial Community Ecology and Physiology
- Time Series Analysis and Forecasting
- Optical measurement and interference techniques
- Marine and coastal ecosystems
- Advanced Electrical Measurement Techniques
- Music and Audio Processing
- Rough Sets and Fuzzy Logic
Institute of Computing Technology
2016-2025
Chinese Academy of Sciences
2016-2025
University of Chinese Academy of Sciences
2010-2025
Research Center for Eco-Environmental Sciences
2020-2025
Isuzu Motors (United States)
2025
Zhejiang Sci-Tech University
2025
Beijing Information Science & Technology University
2025
East China Normal University
2021-2024
Zhengzhou University
2021-2024
Wuxi People's Hospital
2023-2024
Transfer learning aims at improving the performance of target learners on domains by transferring knowledge contained in different but related source domains. In this way, dependence a large number target-domain data can be reduced for constructing learners. Due to wide application prospects, transfer has become popular and promising area machine learning. Although there are already some valuable impressive surveys learning, these introduce approaches relatively isolated way lack recent...
For a target task where labeled data is unavailable, domain adaptation can transfer learner from different source domain. Previous deep methods mainly learn global shift, i.e., align the and distributions without considering relationships between two subdomains within same category of domains, leading to unsatisfying learning performance capturing fine-grained information. Recently, more researchers pay attention Subdomain Adaptation which focuses on accurately aligning relevant subdomains....
To solve the information explosion problem and enhance user experience in various online applications, recommender systems have been developed to model users' preferences. Although numerous efforts made toward more personalized recommendations, still suffer from several challenges, such as data sparsity cold-start problems. In recent years, generating recommendations with knowledge graph side has attracted considerable interest. Such an approach can not only alleviate above mentioned issues...
Abstract Integrated network analysis pipeline (iNAP) is an online for generating and analyzing comprehensive ecological networks in microbiome studies. It implemented two sections, that is, construction analysis, integrates many open‐access tools. Network contains multiple feasible alternatives, including correlation‐based approaches (Pearson's correlation Spearman's rank along with random matrix theory, sparse correlations compositional data) conditional dependence‐based methods (extended...
Graph-based fraud detection approaches have escalated lots of attention recently due to the abundant relational information graph-structured data, which may be beneficial for fraudsters. However, GNN-based algorithms could fare poorly when label distribution nodes is heavily skewed, and it common in sensitive areas such as financial fraud, etc. To remedy class imbalance problem graph-based detection, we propose a Pick Choose Graph Neural Network (PC-GNN short) imbalanced supervised learning...
The rapid proliferation of knowledge graphs (KGs) has changed the paradigm for various AI-related applications. Despite their large sizes, modern KGs are far from complete and comprehensive. This motivated research in graph completion (KGC), which aims to infer missing values incomplete triples. However, most existing KGC models treat triples independently without leveraging inherent valuable information local neighborhood surrounding an entity. To this end, we propose a Relational Graph...
Cold-start problem is still a very challenging in recommender systems. Fortunately, the interactions of cold-start users auxiliary source domain can help recommendations target domain. How to transfer user's preferences from domain, key issue Cross-domain Recommendation (CDR) which promising solution deal with problem. Most existing methods model common preference bridge for all users. Intuitively, since vary user user, bridges different should be different. Along this line, we propose novel...
Pre-training models have shown their power in sequential recommendation. Recently, prompt has been widely explored and verified for tuning after pre-training NLP, which helps to more effectively parameter-efficiently extract useful knowledge from downstream tasks, especially cold-start scenarios. However, it is challenging bring prompt-tuning NLP recommendation, since the tokens of recommendation (i.e., items) are million-level do not concrete explainable semantics, sequence modeling should...
Searching frequent patterns in transactional databases is considered as one of the most important data mining problems and Apriori typical algorithms for this task. Developing fast efficient that can handle large volumes becomes a challenging task due to databases. In paper, we implement parallel algorithm based on MapReduce, which framework processing huge datasets certain kinds distributable using number computers (nodes). The experimental results demonstrate proposed scale well...
Click-through rate (CTR) prediction has been one of the most central problems in computational advertising. Lately, embedding techniques that produce low-dimensional representations ad IDs drastically improve CTR accuracies. However, such learning are data demanding and work poorly on new ads with little logging data, which is known as cold-start problem.
Transfer learning aims at improving the performance of target learners on domains by transferring knowledge contained in different but related source domains. In this way, dependence a large number domain data can be reduced for constructing learners. Due to wide application prospects, transfer has become popular and promising area machine learning. Although there are already some valuable impressive surveys learning, these introduce approaches relatively isolated way lack recent advances...
Default user detection plays one of the backbones in credit risk forecasting and management. It aims at, given a set corresponding features, e.g., patterns extracted from trading behaviors, predicting polarity indicating whether will fail to make required payments future. Recent efforts attempted incorporate attributed heterogeneous information network (AHIN) for extracting complex interactive features users achieved remarkable success on discovering specific default such as fraud, cash-out...
To the same utterance, people's responses in everyday dialogue may be diverse largely terms of content semantics, speaking styles, communication intentions and so on. Previous generative conversational models ignore these 1-to-n relationships between a post to its responses, tend return high-frequency but meaningless responses. In this study we propose mechanism-aware neural machine for response generation. It assumes that there exists some latent responding mechanisms, each which can...
Fraud transactions have been the major threats to healthy development of e-commerce platforms, which not only damage user experience but also disrupt orderly operation market. User behavioral data is widely used detect fraud transactions, and recent works show that accurate modeling intentions in sequences can propel further improvements on performances. However, most existing methods treat each transaction as an independent instance without considering transaction-level interactions...
Though Graph Neural Networks (GNNs) have been successful for fraud detection tasks, they suffer from imbalanced labels due to limited compared the overall userbase. This paper attempts resolve this label-imbalance problem GNNs by maximizing AUC (Area Under ROC Curve) metric since it is unbiased with label distribution. However, on GNN tasks intractable potential polluted topological structure caused intentional noisy edges generated fraudsters. To alleviate problem, we propose decouple...