A5028791879- Recommender Systems and Techniques
- Privacy-Preserving Technologies in Data
- Advanced Graph Neural Networks
- Stochastic Gradient Optimization Techniques
- Topic Modeling
- Cryptography and Data Security
- Domain Adaptation and Few-Shot Learning
- Advanced Bandit Algorithms Research
- Caching and Content Delivery
- Machine Learning in Healthcare
- Human Mobility and Location-Based Analysis
- Adversarial Robustness in Machine Learning
- Fault Detection and Control Systems
- Text and Document Classification Technologies
- Biosensors and Analytical Detection
- Machine Fault Diagnosis Techniques
- Machine Learning and Data Classification
- Image Retrieval and Classification Techniques
- Stock Market Forecasting Methods
- Internet Traffic Analysis and Secure E-voting
- Imbalanced Data Classification Techniques
- Advanced biosensing and bioanalysis techniques
- Advanced Battery Technologies Research
- Data Stream Mining Techniques
- Mental Health via Writing
Zhejiang University
2013-2025
Wenzhou Medical University
2020-2025
China Institute of Veterinary Drug Control
2024-2025
Aerospace Information Research Institute
2025
Chinese Academy of Sciences
2025
Sun Yat-sen Memorial Hospital
2025
Sun Yat-sen University
2025
Zhejiang University of Science and Technology
2015-2024
China Agricultural University
2018-2024
Western University
2021-2024
We present, GEM, the first heterogeneous graph neural network approach for detecting malicious accounts at Alipay, one of world's leading mobile cashless payment platform. Our approach, inspired from a connected subgraph adaptively learns discriminative embeddings account-device graphs based on two fundamental weaknesses attackers, i.e. device aggregation and activity aggregation. For consists various types nodes, we propose an attention mechanism to learn importance different while using...
We present, GeniePath, a scalable approach for learning adaptive receptive fields of neural networks defined on permutation invariant graph data. In we propose an path layer consists two complementary functions designed breadth and depth exploration respectively, where the former learns importance different sized neighborhoods, while latter extracts filters signals aggregated from neighbors hops away. Our method works in both transductive inductive settings, extensive experiments compared...
In order to address the data sparsity problem in recommender systems, recent years, Cross-Domain Recommendation (CDR) leverages relatively richer information from a source domain improve recommendation performance on target with sparser information. However, each of two domains may be certain types (e.g., ratings, reviews, user profiles, item details, and tags), thus, if we can leverage such well, it is possible both simultaneously (i.e., dual-target CDR), rather than single only. To this...
To address the long-standing data sparsity problem in recommender systems (RSs), cross-domain recommendation (CDR) has been proposed to leverage relatively richer information from a domain improve performance sparser domain. Although CDR extensively studied recent years, there is lack of systematic review existing approaches. fill this gap, paper, we provide comprehensive approaches, including challenges, research progress, and prospects. Specifically, first summarize approaches into four...
Machine prognosis is a significant part of condition-based maintenance and intends to monitor track the time evolution fault so that can be performed or task terminated avoid catastrophic failure. A new prognostic method developed in this paper using adaptive neuro-fuzzy inference systems (ANFISs) high-order particle filtering. The ANFIS trained via machine historical failure data. its modeling noise constitute an <i xmlns:mml="http://www.w3.org/1998/Math/MathML"...
This paper proposes a novel approach for machine health condition prognosis based on neuro-fuzzy systems (NFSs) and Bayesian algorithms. The NFS, after training with data, is employed as prognostic model to forecast the evolution of fault state time. An online update scheme developed basis probability density function (PDF) NFS residuals between actual predicted data. estimation algorithms adopt model's data prior information in combination measurements degree belief forecasting estimations....
Cross-Domain Recommendation (CDR) and Cross-System Recommendations (CSR) are two of the promising solutions to address long-standing data sparsity problem in recommender systems. They leverage relatively richer information, e.g., ratings, from source domain or system improve recommendation accuracy target system. Therefore, finding an accurate mapping latent factors across domains systems is crucial enhancing accuracy. However, this a very challenging task because complex relationships...
The conventional single-target Cross-Domain Recommendation (CDR) only improves the recommendation accuracy on a target domain with help of source (with relatively richer information). In contrast, novel dual-target CDR has been proposed to improve accuracies both domains simultaneously. However, faces two new challenges: (1) how generate more representative user and item embeddings, (2) effectively optimize user/item embeddings each domain. To address these challenges, in this paper, we...
Cross Domain Recommendation (CDR) has been popularly studied to alleviate the cold-start and data sparsity problem commonly existed in recommender systems. CDR models can improve recommendation performance of a target domain by leveraging other source domains. However, most existing assume information directly 'transfer across bridge', ignoring privacy issues. To solve concern CDR, this paper, we propose novel two stage based privacy-preserving framework (PriCDR). In first stage, methods,...
Abstract Biocatalytic cascades are challenging to operate in homogeneous solution, where diffusional mass transport hinders efficient communication between the reactive components. There is great interest developing devices perform such transformations confined environments, which increase efficiency of cascaded process by generating high local concentrations species. Herein, a bioreactor‐nanozyme assembly introduced for aerobic oxidation N ‐hydroxy‐ l ‐arginine (NOHA) citrulline presence...
Sequential Recommendation (SR) characterizes evolving patterns of user behaviors by modeling how users transit among items. However, the short interaction sequences limit performance existing SR. To solve this problem, we focus on Cross-Domain (CDSR) in paper, which aims to leverage information from other domains improve sequential recommendation a single domain. Solving CDSR is challenging. On one hand, retain domain preferences as well integrate cross-domain influence remains an essential...
Cross-Domain Recommendation (CDR) has been popularly studied to utilize different domain knowledge solve the data sparsity and cold-start problem in recommender systems. In this paper, we focus on Review-based Non-overlapped (RNCDR) problem. The is commonly-existed challenging due two main aspects, i.e, there are only positive user-item ratings target no overlapped user across domains. Most previous CDR approaches cannot RNCDR well, since (1) they effectively combine review with other...
Sequential Cross-Domain Recommendation (CDR) has been popularly studied to utilize different domain knowledge and users' historical behaviors for the next-item prediction. In this paper, we focus on cross-domain sequential recommendation problem. This commonly exist problem is rather challenging from two perspectives, i.e., implicit user rating sequences are difficult in modeling users/items domains mostly non-overlapped. Most previous CDR approaches cannot solve well, since (1) they...
Session-based Recommendation aims at predicting the next interacted item based on short anonymous behavior sessions. However, existing solutions neglect to model two inherent properties of sequential representing distributions, i.e., hierarchy structures resulted from popularity and collaborations in both intra- inter-session. Tackling with these factors same time is challenging. On one hand, traditional Euclidean space utilized previous studies fails capture due a restricted representation...
With the rapid development of Internet and Web techniques, Cross-Domain Recommendation (CDR) models have been widely explored for resolving data-sparsity cold-start problem. Meanwhile, most CDR should utilize explicit domain-shareable information (e.g., overlapped users or items) knowledge transfer across domains. However, this assumption may not be always satisfied since items are non-overlapped in real practice. The performance many previous works will severely impaired when these...
Internet companies are facing the need for handling large-scale machine learning applications on a daily basis and distributed implementation of algorithms which can handle extra-large-scale tasks with great performance is widely needed. Deep forest recently proposed deep framework uses tree ensembles as its building blocks it has achieved highly competitive results various domains tasks. However, not been tested extremely In this work, based our parameter server system, we developed version...