A5002597419- Recommender Systems and Techniques
- Advanced Graph Neural Networks
- Topology Optimization in Engineering
- Composite Structure Analysis and Optimization
- Advanced Multi-Objective Optimization Algorithms
- Caching and Content Delivery
- Privacy-Preserving Technologies in Data
- Topic Modeling
- Neural Networks and Applications
- FinTech, Crowdfunding, Digital Finance
- Data Stream Mining Techniques
- Consumer Market Behavior and Pricing
- Machine Learning and ELM
- Domain Adaptation and Few-Shot Learning
- Open Source Software Innovations
- Advanced Numerical Analysis Techniques
- Statistical and numerical algorithms
- Bayesian Methods and Mixture Models
- Extracellular vesicles in disease
- Human Mobility and Location-Based Analysis
- Data Visualization and Analytics
- Mobile Crowdsensing and Crowdsourcing
- Composite Material Mechanics
- Nanopore and Nanochannel Transport Studies
- Advanced Bandit Algorithms Research
Hunan University of Traditional Chinese Medicine
2022
Hunan Provincial Science and Technology Department
2022
Southwest Jiaotong University
2022
Zhejiang Financial College
2019-2021
Dalian University of Technology
2013-2015
University of Hong Kong
2010
Hong Kong University of Science and Technology
2010
In the present work, we intend to demonstrate how do topology optimization in an explicit and geometrical way. To this end, a new computational framework for structural based on concept of moving morphable components is proposed. Compared with traditional pixel or node point-based solution framework, proposed paradigm can incorporate more geometry mechanical information into directly therefore render process flexibility. It also has great potential reduce burden associated substantially....
The recently explosive growth of Super Apps brings great convenience to people's daily life by providing a wide variety services through mini-programs, including online shopping, travel, finance, and so on. Due the considerable gap between various scenarios, restriction effective information transfer sharing severely blocks efficient delivery services, potentially affecting user's app experience. To deeply understand users' needs, we propose SupKG, commonsense knowledge graph towards APP...
To help merchants/customers to provide/access a variety of services through miniapps, online service platforms have occupied critical position in the effective content delivery, which how recommend items new domain launched by provider for customers has become more urgent. However, non-negligible gap between source and diversified target domains poses considerable challenge cross-domain recommendation systems, often leads performance bottlenecks industrial settings. While entity graphs...
Federated learning (FL) is a popular way of edge computing that doesn't compromise users' privacy. Current FL paradigms assume data only resides on the edge, while cloud servers perform model averaging. However, in real-life situations such as recommender systems, server has ability to store historical and interactive features. In this paper, our proposed Edge-Cloud Collaborative Knowledge Transfer Framework (ECCT) bridges gap between cloud, enabling bi-directional knowledge transfer both,...
Despite that path-based and embedding-based models with knowledge graphs (KGs) achieve better recommendation performance compared other deep learning based methods, such improvement is limited due to a lack of modeling user's dynamic interest. To address this issue, we explore principled model provide semantic understanding each item in historical interest sequence KGs. Specifically, propose multi-granularity method, which on knowledge-enhanced path mining fluctuation signal discovery,...
Line-commutated converter based high-voltage direct-current (LCC-HVDC) transmission systems are prone to subsequent commutation failure (SCF), which consequently leads the forced blocking of HVDC links, affecting operation power system. An accurate (CF) identification is fairly vital prevention SCF. However, existing CF methods cause misjudge or detection lag, can limit effect SCF mitigation strategy. In addition, earlier approaches suppress do not clarify key factor that determines...
Mobile payment such as Alipay has been widely used in our daily lives. To further promote the mobile activities, it is important to run marketing campaigns under a limited budget by providing incentives coupons, commissions merchants. As result, incentive optimization key maximizing commercial objective of campaign. With analyses online experiments, we found that transaction network can subtly describe similarity merchants' responses different incentives, which great use problem. In this...
Clustering using the Hilbert Schmidt independence criterion (CLUHSIC) is a recent clustering algorithm that maximizes dependence between cluster labels and data observations according to (HSIC). It unique in structure information on outputs can be easily utilized process. However, while choice of loss function known very important supervised learning with structured outputs, we will show this paper CLUHSIC implicitly often inappropriate zero-one loss. We propose an extension called CLUHSICAL...
Recently, modeling temporal patterns of user-item interactions have attracted much attention in recommender systems. We argue that existing methods ignore the variety user behaviors. define subset behaviors are irrelevant to target item as noises, which limits performance target-related time cycle and affect recommendation performance. In this paper, we propose Denoising Time Cycle Modeling (DiCycle), a novel approach denoise select highly related item. DiCycle is able explicitly model...
In Federated Learning (FL), the data in each client is typically assumed fixed or static. However, often comes an incremental manner real-world applications, where domain may increase dynamically. this work, we study catastrophic forgetting with heterogeneity Incremental (FIL) scenarios edge clients lack enough storage space to retain full data. We propose employ a simple, generic framework for FIL named Re-Fed, which can coordinate cache important samples replay. More specifically, when new...
Online service platforms offering a wide range of services through miniapps have become crucial for users who visit these with clear intentions to find they are interested in. Aiming at effective content delivery, cross-domain recommendation introduced learn high-quality representations by transferring behaviors from data-rich scenarios. However, methods overlook the impact decision path that take when conduct behaviors, is, ultimately exhibit different based on various intents. To this end,...
The rapid evolution of deep learning and large language models has led to an exponential growth in the demand for training data, prompting development Dataset Distillation methods address challenges managing datasets. Among these, Matching Training Trajectories (MTT) been a prominent approach, which replicates trajectory expert network on real data with synthetic dataset. However, our investigation found that this method suffers from three significant limitations: 1. Instability generated by...
Facing the challenges of sparsity and long tail in thousands Mini-apps recommendation scenarios deployed on Alipay platform, there is a great need for simple, effective easy-to-deploy industrial solution. To address this issue, we follow strategy ‘divide conquer’ propose crowd-based model by using D eterminantal P oint rocesse s C rowd-wise M ixture- o f- E xperts (DPPs-CMoE). Specifically, under guidance DPPs-based prototypical tags, user profiling space sequentially divided into multiple...
In this paper, we highlight that both conformity and risk preference matter in making fund investment decisions beyond personal interest seek to jointly characterize these aspects a disentangled manner. Consequently, develop novel Multi-granularity Graph Disentangled Learning framework named MGDL effectively perform intelligent matching of products. Benefiting from the well-established graph attention module, multi-granularity user representations are derived historical behaviors separately...
The first provably efficient algorithm for learning graph neural networks (GNNs) with one hidden layer node information convolution is provided in this paper. Two types of GNNs are investigated, depending on whether labels attached to nodes or graphs. A comprehensive framework designing and analyzing convergence GNN training algorithms developed. proposed applicable a wide range activation functions including ReLU, Leaky Sigmod, Softplus Swish. It shown that the guarantees linear rate...
Deep Candidate Generation plays an important role in large-scale recommender systems. It takes user history behaviors as inputs and learns item latent embeddings for candidate generation. In the literature, conventional methods suffer from two problems. First, a has multiple to reflect various interests, such number is fixed. However, taking into account different levels of activeness, fixed interest sub-optimal. For example, less active users, they may need fewer represent their interests...