A5044163944- Recommender Systems and Techniques
- Advanced Graph Neural Networks
- Privacy-Preserving Technologies in Data
- Advanced Bandit Algorithms Research
- Visual Attention and Saliency Detection
- Stochastic Gradient Optimization Techniques
- Radiomics and Machine Learning in Medical Imaging
- Non-Destructive Testing Techniques
- Topic Modeling
- Magnetic Properties and Applications
- Welding Techniques and Residual Stresses
- Caching and Content Delivery
- Transportation and Mobility Innovations
Zhejiang University of Science and Technology
2023-2024
Zhejiang University
2022-2024
China National Petroleum Corporation (China)
2019
Cross-Domain Recommendation (CDR) aims to solve the data sparsity problem by integrating strengths of different domains. Though researchers have proposed various CDR methods effectively transfer knowledge across domains, they fail address following key issues, i.e., (1) cannot model high-order correlations among users and items in every single domain obtain more accurate representations; (2) To tackle above we propose a novel Intra Inter Domain HyperGraph Convolutional Network (II-HGCN)...
Federated learning (FL) has garnered considerable interest for its capability to learn from decentralized data sources. Given the increasing application of FL in decision-making scenarios, addressing fairness issues across different sensitive groups (e.g., female, male) is crucial. Current research typically focus on facilitating at each client's (local fairness) or within entire dataset all clients (global fairness). However, existing approaches that exclusively either global local fail...
With the growing privacy concerns in recommender systems, recommendation unlearning, i.e., forgetting impact of specific learned targets, is getting increasing attention. Existing studies predominantly use training data, model inputs, as unlearning target. However, we find that attackers can extract private information, gender, race, and age, from a trained even if it has not been explicitly encountered during training. We name this unseen information attribute treat To protect sensitive...
Recommender systems are typically biased toward a small group of users, leading to severe unfairness in recommendation performance, i.e., User-Oriented Fairness (UOF) issue. Existing research on UOF exhibits notable limitations two phases models. In the training phase, current methods fail tackle root cause issue, which lies unfair process between advantaged and disadvantaged users. evaluation metric lacks ability comprehensively evaluate varying cases unfairness. this paper, we aim address...
The development of coiled tubing operation technology has continuously improved with the higher requirements on-line detection technology. current field weak magnetic system calculates gradient signal for improving locating precision. Quantitative identification defect shape, based on and leakage flux systems, is a key issue to evaluate whether can continue be used. 2-D finite-element models different typical shapes were established, changes magnetic-field intensity are calculated. influence...
Recommender systems are typically biased toward a small group of users, leading to severe unfairness in recommendation performance, i.e., User-Oriented Fairness (UOF) issue.The existing research on UOF is limited and fails deal with the root cause issue: learning process between advantaged disadvantaged users unfair.To tackle this issue, we propose an In-processing User Constrained Dominant Sets (In-UCDS) framework, which general framework that can be applied any backbone model achieve...
Recommender systems are fundamental information filtering techniques to recommend content or items that meet users’ personalities and potential needs. As a crucial solution address the difficulty of user identification unavailability historical information, session-based recommender provide recommendation services only rely on behaviors in current session. However, most existing studies not well-designed for modeling heterogeneous capturing relationships between them practical scenarios. To...
The emergence of ChatGPT marks the arrival large language model (LLM) era. While LLMs demonstrate their power in a variety fields, they also raise serious privacy concerns as users' queries are sent to provider. On other side, deploying LLM on user's device will leak all data. Existing methods based secure multiparty computation (MPC) managed protect both parameters and user queries. However, require gigabytes data transfer several minutes generate just one token, making them impractical for...
Federated Learning (FL) employs a training approach to address scenarios where users' data cannot be shared across clients. Achieving fairness in FL is imperative since inherently geographically distributed among diverse user groups. Existing research on predominantly assumes access the entire data, making direct transfer challenging. However, limited existing does not effectively two key challenges, i.e., (CH1) Current methods fail deal with inconsistency between fair optimization results...
Recommender systems are fundamental information filtering techniques to recommend content or items that meet users' personalities and potential needs. As a crucial solution address the difficulty of user identification unavailability historical information, session-based recommender provide recommendation services only rely on behaviors in current session. However, most existing studies not well-designed for modeling heterogeneous capturing relationships between them practical scenarios. To...