- Advanced Graph Neural Networks
- Topic Modeling
- Advanced Algorithms and Applications
- Cryptography and Data Security
- Data Mining Algorithms and Applications
- Cognitive Computing and Networks
- Rough Sets and Fuzzy Logic
- Data Quality and Management
- Face and Expression Recognition
- Cloud Data Security Solutions
- Domain Adaptation and Few-Shot Learning
- Access Control and Trust
- Caching and Content Delivery
- Recommender Systems and Techniques
- Imbalanced Data Classification Techniques
- Distributed and Parallel Computing Systems
- Security in Wireless Sensor Networks
- Data Stream Mining Techniques
- Indoor and Outdoor Localization Technologies
- Fault Detection and Control Systems
- Geoscience and Mining Technology
- Image and Video Stabilization
- Natural Language Processing Techniques
- Machine Learning and ELM
- Gait Recognition and Analysis
Xi'an Jiaotong University
1992-2024
Beijing Institute of Big Data Research
2024
University of Regina
2004
Recent years have witnessed the rapid development in research topic of WiFi sensing that automatically senses human with commercial devices. This work falls into two major categories, i.e., activity recognition and indoor localization. The former utilizes devices to recognize daily activities such as smoking, walking, dancing. latter one, localization, can be used for navigation, location-based services, through-wall surveillance. key rationale behind this type is people behaviors influence...
Knowledge graph completion (KGC) aims to study the embedding representation solve incompleteness of knowledge graphs (KGs). Recently, convolutional networks (GCNs) and attention (GATs) have been widely used in KGC tasks by capturing neighbor information entities. However, Both GCNs GATs based models their limitations, best method is analyze neighbors each entity (pre-validating), while this process prohibitively expensive. Furthermore, quality embeddings can affect aggregation (message...
Recently, an enormous amount of research has emerged on multimodal knowledge graph completion (MKGC), which seeks to extract from data and predict the most plausible missing facts complete a given (MKG). However, existing MKGC approaches largely ignore that visual information may introduce noise lead uncertainty when adding them traditional KG embeddings due contribution each associated image entity is different in diverse link scenarios. Moreover, treating triple independently learning...
Recently, a large amount of work has emerged for knowledge graph completion (KGC), which aims to reason over known facts and infer the missing links. Meanwhile, contrastive learning been applied KGC tasks, can improve representation quality entities relations. However, existing approaches tend their performance with high-dimensional embeddings complex models, make them suffer from storage space high training costs. Furthermore, loss single positive sample learns little structural semantic...
The efficiency of mining association rules is an important field KDD. algorithm Apriori a classical in rules. It breadth first search on the lattice space itemsets. Though it makes use anti-monotone itemsets to reduce searching breadth, algorithmic complexity time still exponential quantity. In this article, concepts generation and ordinal tree are introduced. dynamic description relation itemsets, vegetal ability described by generation. Through study rules, conclusion that all frequent not...
This paper discusses the current status of research about mining association rules in a database, which points out shortcoming classical priori's algorithm, and presents some theorems based on reducing records larger database. It also applies theory fuzzy sets to processing data traps network management. According those sets, we have designed AprioriFuzzy algorithm mine Through performance is analyzed evaluated this paper, can save time 30% effectively. has been implemented PC Visual C++6.0...
Recently, graph convolutional networks (GCNs) and attention (GATs) have been widely used in knowledge completion (KGC), which aims to solve the incompleteness of graphs (KGs). However, GCNs based KGC models tend stack redundant information when one entity several noisy neighbors. GATs may focus on certain neighbors lead weakening structural information. Furthermore, relations KGs their specific semantics should be considered aggregating neighbor (message passing). To address above...
Logic key hierarchy is a leading scheme for distribution in secure multicast. An improved logic scheme, based on one-way hash function, proposed. Besides its simplicity, the can reduce server's average cost by about 1/3. Moreover, batch update algorithm this also presented, and performance evaluated through experiments. Compared with it shows presented paper server at least
The main stream of research in data mining (or knowledge discovery databases) focuses on algorithms and automatic or semi-automatic processes for discovering hidden data. In this paper, we adopt a more general goal oriented view mining. Data is regarded as field study covering the theories, methodologies, techniques, activities with new useful knowledge. One its objectives to design implement systems. A miner solves problems manually, semi-automatically by using such However, there lack...
摘要: 在高维空间中,分类超平面倾向于通过原点,即不需要偏置(b)。为了研究在-SVM分类问题中是否需要b,该文提出了无(b)的-SVM的对偶优化问题并给出了其优化问题求解方法。该方法通过有效集策略将对偶优化问题转化为等式约束子优化问题,然后通过拉格朗日乘子法将子优化问题转化为线程方程组来求解。实验表明偏置(b)的存在会降低-SVM的泛化性能,-SVM只能得到无(b) -SVM的次优解。 关键词: -支持向量机 / 偏置 泛化性能 有效集
PDF HTML阅读 XML下载 导出引用 引用提醒 无偏置支持向量回归优化问题 DOI: 10.3724/SP.J.1001.2012.04150 作者: 作者单位: 作者简介: 通讯作者: 中图分类号: 基金项目: 国家自然科学基金(61173040) Support Vector Regression Optimization Problem without Bias Author: Affiliation: Fund Project: 摘要 | 图/表 访问统计 参考文献 相似文献 引证文献 资源附件 文章评论 摘要:为了研究偏置对支持向量回归(support vector regression,简称SVR)问题泛化性能的影响,首先提出了无偏置SVR(NBSVR)的优化问题及其对偶问题.推导出了NBSVR 优化问题全局最优解的必要条件,然后证明了SVR 的对偶问题只能得到NBSVR 对偶问题的次优解.同时提出了NBSVR 的有效集求解算法,并证明了它是线性收敛的.基于21...
Role-based encryption is a new cryptographic primitive that enables the role-based access control (RBAC) model for encrypted data in cloud storage environments. Compared with some other technologies, such as attribute-based (ABE), it can greatly relieve owners from heavy burden to define and manage policies. In this paper, we present generic construction of inner-product (IPE) revocation encryption, following two significant features: 1) first adaptively secure scheme standard model; 2) more...
Role-based encryption is a new cryptographic primitive that enables the role-based access control (RBAC) model for encrypted data in cloud storage environments. Compared with some other technologies, such as attribute-based (ABE), it can greatly relieve owners from heavy burden to define and manage policies. In this paper, we present generic construction of inner-product (IPE) revocation encryption, following two significant features: 1) first adaptively secure scheme standard model; 2) more...
Certificate revocation is an outstanding problem in PKI. This paper extends Naor's scheme of dynamic hash tree order to optimize performance. Set revoked certificates divided into groups. In each group, proofs for certificate status are computed by using one-way accumulator, while all groups still organized tree. The main advantage the proposed that it can adjust traffic between CA-to-directory and directory-to-user according update rate query applications, thus remarkably reduce overall...
Based on the wavelet decomposition and conditions of support vector kernel function, Morlet function for machine (SVM) is proposed, which a kind approximately orthonormal function. This can simulate almost any curve in quadratic continuous integral space, thus it enhances generalization ability SVM. According to regularization theory, Least squares (LS-MWSVM) proposed simplify process MWSVM. The LS-MWSVM then applied regression analysis or classifying. Experiment results show that...
Multi-hop reasoning is an effective and explainable approach to predicting missing facts in Knowledge Graphs (KGs). It usually adopts the Reinforcement Learning (RL) framework searches over KG find evidential path. However, there are few RL based methods Forests (KFs), existing transfer multi-hop KF. this way often leads a long path with lot of redundant information. And tend capture vector representations from local structures, which will lead get meaningless paths. In order solve these two...
Integrating lexicon knowledge into character-based methods can improve the performance of neural network models for Chinese named entity recognition (NER). For example, Lattice LSTM [1]and WC-LSTM [2] perform well on several public NER datasets. However, directed acyclic graph (DAG) structure makes lattice challenging to train minibatch. In addition, and only incorporate word-level semantics representation first or last character in each word. The inside characters that word contain are...
This article presents the architectural design and implementation of a Lisp machine Lisp_M1 developed at Xian Jiaotong University in China. Approaches adopted by to support execution Programs are discussed. Techniques used implement presented. A speed comparison between some conventional computers special purpose machines is also given.