A5100636815- Advanced Differential Equations and Dynamical Systems
- Algebraic Geometry and Number Theory
- Advanced Graph Neural Networks
- Polynomial and algebraic computation
- Recommender Systems and Techniques
- Advanced Topics in Algebra
- Topic Modeling
- Privacy-Preserving Technologies in Data
- Computational Drug Discovery Methods
- Complex Network Analysis Techniques
- Mathematical Dynamics and Fractals
- Imbalanced Data Classification Techniques
- graph theory and CDMA systems
- Stochastic Gradient Optimization Techniques
- Speech and dialogue systems
- Cybercrime and Law Enforcement Studies
- Advanced Decision-Making Techniques
- Finite Group Theory Research
- Geometric and Algebraic Topology
- Spam and Phishing Detection
- Rings, Modules, and Algebras
- Evaluation and Optimization Models
- Natural Language Processing Techniques
- Commutative Algebra and Its Applications
- Soil, Finite Element Methods
Xi'an Jiaotong University
2015-2024
State Key Laboratory of Electrical Insulation and Power Equipment
2024
University of Bonn
2024
Beijing Proteome Research Center
2021
Zhejiang Financial College
2018-2020
Abstract Combination therapy has shown an obvious efficacy on complex diseases and can greatly reduce the development of drug resistance. However, even with high-throughput screens, experimental methods are insufficient to explore novel combinations. In order search space combinations, there is urgent need develop more efficient computational predict recent decades, machine learning (ML) algorithms have been applied improve predictive performance. The object this study introduce discuss...
Differentially private collaborative filtering is a challenging task, both in terms of accuracy and speed. We present simple algorithm that provably differentially private, while offering good performance, using novel connection differential privacy to Bayesian posterior sampling via Stochastic Gradient Langevin Dynamics. Due its simplicity the lends itself efficient implementation. By careful systems design by exploiting power law behavior data maximize CPU cache bandwidth we are able...
The article revisits birational and biregular automorphisms of the Hilbert scheme points on a K3 surface from perspective derived categories. Under assumption that is generic, involutions induced by autoequivalences category underlying are characterized.
Fraudulent claim detection is one of the greatest challenges insurance industry faces. Alibaba's return-freight insurance, providing return-shipping postage compensations over product return on e-commerce platform, receives thousands potentially fraudulent claims everyday. Such deliberate abuse policy could lead to heavy financial losses. In order detect and prevent claims, we developed a novel data-driven procedure identify groups organized fraudsters, major contributions losses, by...
Recently, Graph Neural Network (GNN) has achieved remarkable progresses in various real-world tasks on graph data, consisting of node features and the adjacent information between different nodes. High-performance GNN models always depend both rich complete edge graph. However, such could possibly be isolated by data holders practice, which is so-called isolation problem. To solve this problem, paper, we propose VFGNN, a federated learning paradigm for privacy-preserving classification task...
Recently, in order to take a preemptive opportunity the mobile economy, Internet companies conduct thousands of marketing campaigns every day, promote their products and services. In scenario, one fundamental issues is audience expansion task for campaigns. Given set seed users, aims seek more users (audiences), who are similar seeds will finish business goal targeted campaign (ie convert). However, problem challenging three aspects. First, company run hundreds serve massive day. The...
Abstract We determine the group of all Fourier–Mukai type autoequivalences Kuznetsov components smooth complex cubic threefolds, and provide yet another proof for version categorical Torelli theorem threefolds.
Collaborative filtering, especially latent factor model, has been popularly used in personalized recommendation. Latent model aims to learn user and item factors from user-item historic behaviors. To apply it into real big data scenarios, efficiency becomes the first concern, including offline training online recommendation efficiency. In this paper, we propose a D istributed C ollaborative H ashing ( DCH ) which can significantly improve both efficiencies. Specifically, distributed learning...
In Baker and Lorscheid's paper, they introduce a new hyperstructure: the polynomial hyperstructure Poly$(\mathbb{F})$ over hyperfield $\mathbb{F}$. this work, author focuses on associativity of hypermultiplications in those hyperstructures gives elementary propositions. The also shows examples hyperfields with non-associative hypermultiplications. Then, he proves that though hypermultiplication Poly$(\mathbb{T})$ is associative for linear polynomials, it not general. Moreover, if...
Multi-modal knowledge graph completion (MMKGC) aims to automatically discover new triples in the given multi-modal graphs (MMKGs), which is achieved by collaborative modeling structural information concealed massive and features of entities. Existing methods tend focus on crafting elegant entity-wise fusion strategies, yet they overlook utilization multi-perspective within modalities under diverse relational contexts. To address this issue, we introduce a novel MMKGC framework with Mixture...
NFT는 전통적인 예술 창작 및 거래 모델을 뒤집고 창작자와 수집가를 위한 탈중앙 플랫폼을 구축하며 새로운 디지털 유통 형태를 제공한다. 개인 이미지 PFP를 중심으로 한 NFT 작품은 소셜 미디어의 정체성과 지위를 상징하며 브랜드 구축에서 중요한 역할을 한다. 이 연구는 예술의 시각 인지 요소 통해 예술가와 수집가의 시장 발전을 이론적 지원을 제공하며 시각적 디자인 혁신과 실천을 촉진고자 먼저 사전 조사와 문헌 검토를 PFP의 개념과 PFP 작품의 단계를 분석한다. 둘째, 질적 연구 방법을 사용하여 36명의 소비자를 대상으로 반구조화된 인터뷰를 진행하고, 근거이론을 인터뷰 데이터를 3단계로 인코딩하며, 제안한다. 셋째, 구조 방정식 모델 SEM과 SPSS를 구축된 정량적으로 분석하고 검증한다. 또한 모델에서 제안한 요소의 상호 관계를 각 요소가 인지에 어떤 영향을 미치는지 설명한다. 결과, 예술품의 인지에서 색상, 도형, 형태, 문화, 스타일 창의성의 6가지 핵심 함을...
Massive users' online adoption behaviors were recorded thanks to the various emerging web services such as Facebook, Twitter, G+, Netflix and so on. Two key factors that affect are social selection influence. Understanding underlying each behavior can potentially help service providers gain much more insights into their users improve predictive power. In this paper, we try answer (1) How do roles of influence play in a user-level adoption? (2) Capturing those benefit modeling prediction or...
In this paper we describe an algorithm for predicting the websites at risk in a long range hacking activity, while jointly inferring provenance and evolution of vulnerabilities on over continuous time. Specifically, use hazard regression with time-varying additive function parameterized generalized linear form. The activation coefficients each feature are continuous-time functions constrained total variation penalty inspired by campaigns. We show that optimal solution is 0th order spline...
We determine the group of all Fourier-Mukai type autoequivalences Kuznetsov components smooth complex cubic threefolds, and provide yet another proof for version categorical Torelli theorem threefolds.
The author introduces a conjecture about Makar-Limanov invariants of affine unique factorization domains over field characteristic zero. Then the finds that does not always hold when $\mathbbm{k}$ is algebraically closed and gives some examples where holds.
In this paper, the author introduces hyperfields and give some facts about roots multiplicities of polynomials over hyperfield based on \cite{2} \cite{3}. Then he tests sharpness an inequality in Baker's former work behavior under homomorphisms between show that is not sharp natural $\mathbb{C}\rightarrow\mathbb{P}$ $\mathbb{C}\rightarrow\mathbb{V}$ while it $\mathbb{C}\rightarrow\mathbb{K}$, $\mathbb{R}\rightarrow\mathbb{S}$ $\mathbb{R}\rightarrow\mathbb{T}$ according to previous Baker Lorscheid.
Many classical fairy tales, fiction, and screenplays leverage dialogue to advance story plots establish characters. We present the first study explore whether machines can understand generate in stories, which requires capturing traits of different characters relationships between them. To this end, we propose two new tasks including Masked Dialogue Generation Speaker Recognition, i.e., generating missing turns predicting speakers for specified turns, respectively. build a dataset DialStory,...