A5100411265- Complex Network Analysis Techniques
- Advanced Image and Video Retrieval Techniques
- Advanced Thermodynamics and Statistical Mechanics
- Advanced Graph Neural Networks
- Stochastic processes and statistical mechanics
- Topic Modeling
- Video Surveillance and Tracking Methods
- Anomaly Detection Techniques and Applications
- Cloud Computing and Resource Management
- Neural Networks and Applications
- Visual Attention and Saliency Detection
- Human Pose and Action Recognition
- Advanced Computational Techniques and Applications
- Data Stream Mining Techniques
- Adversarial Robustness in Machine Learning
- Model Reduction and Neural Networks
- Opinion Dynamics and Social Influence
- Machine Learning and ELM
- Multimodal Machine Learning Applications
- Graph Theory and Algorithms
- Markov Chains and Monte Carlo Methods
- Advanced Algorithms and Applications
- Neural dynamics and brain function
- Industrial Technology and Control Systems
- Image and Video Stabilization
Shanghai Polytechnic University
2019-2025
Beijing Institute of Technology
2024
University of Massachusetts Amherst
2018-2023
iDEAL Technology (United States)
2022
Twitter (United States)
2021-2022
University of California, San Francisco
2022
Northeastern University
2015-2022
Chinese Institute for Brain Research
2022
Beijing Normal University
2022
Dalian University
2022
Many massive web and communication network applications create data which can be represented as a sequential stream of edges. For example, conversations in telecommunication or messages social Such streams are typically very large, because the large amount underlying activity such networks. An important application these domains is to determine frequently occurring dense structures graph stream. In general, we would like frequent patterns interactions. We introduce model for pattern mining...
Numerous problems in many fields can be solved effectively through the approach of modeling by complex network analysis. Finding key nodes is one most important and challenging In previous studies, methods have been proposed to identify nodes. However, they rely mainly on a limited field local information, lack large-scale access global are also usually NP-hard. this paper, novel entropy mutual information-based centrality (EMI) proposed, which attempts capture far wider range greater...
Key nodes are similar to important hubs in a network structure, which can directly determine the robustness and stability of network. By effectively identifying protecting these critical nodes, be improved, making it more resistant external interference attacks. There various topology analysis methods for given network, but key node identification often focus on either local attributes or global attributes. Designing an algorithm that combines both improve accuracy identification. In this...
As the manufacturing industry develops towards high precision and intelligence, CNC machine tools play an important role in production. The occurrence of failures not only reduces production efficiency but also increases costs. In order to improve accuracy fault prediction, this paper establishes integrated learning model for tool prediction by stacking ensemble algorithms combining decision trees, support vector (SVM), random forests other algorithms. fault-related features are optimized...
Collective classification in relational data has become an important and active research topic the last decade. It exploits dependencies of instances a network to improve predictions. Related applications include hyperlinked document classification, social analysis collaboration analysis. Most traditional collective models mainly study scenario that there exists large amount labeled examples (labeled nodes). However, many real-world applications, are extremely difficult obtain. For example,...
WeChat is a popular mobile instant messenger (MIM) in China with hundreds of millions users. Recent literature suggested that MIMs can support users to "dwell" together their close relationships by constantly exchanging tidbits lives. In this paper, we present an interview study investigate why and how people use WeChat, particularly focusing on its three novel opportunistic social features: Shake, Drift Bottle, Look Around. Drawing from 25 interviews, our results suggest usage was motivated...
Recognizing the identities of people in everyday photos is still a very challenging problem for machine vision, due to issues such as non-frontal faces, changes clothing, location, lighting. Recent studies have shown that rich relational information between same photo can help recognizing their identities. In this work, we propose model sequence prediction task. At core our work novel recurrent network architecture, which instances labels and appearance are modeled jointly. addition cues,...
In this paper, we examine the problem of node re-identification from anonymized graphs. Typical graphs encountered in real applications are massive and sparse. will show that sparse have certain theoretical properties which make them susceptible to attacks. We design a systematic way exploit these order construct re- identification signatures, also known as characteristic vectors. These signatures property they extremely robust perturbations, especially for Our results even low levels...
In this paper, a hybrid model based on sooty tern optimization algorithm (STOA) is proposed to optimize the parameters of support vector machine (SVM) and identify best feature sets simultaneously. Feature selection an essential process data preprocessing, it aims find most relevant subset features. recent years, has been applied in many practical domains intelligent systems. The application SVM fields proved its effectiveness classification tasks various types. Its performance mainly...
Presto is an open source distributed query engine used widely at Facebook, Uber, Twitter, Pinterest, and many other internet companies. Since sourced in 2013, the community has made several rounds of design implementations, to support a variety use cases, including interactive analytics, real time reporting dashboard, ETL workloads, A/B testing, monitoring alerts, etc. In this paper, we'd like introduce some most important features performance improvements recent years, which enables...
Recent progress in distributed robotics and low power embedded systems has led to development of mobile sensor networks. Controlled mobility, moving sensors intentionally, enables a new set possibilities wireless networks facilitates many applications signal processing areas such as target tracking. In this paper we consider the problem tracking using three that measure received strength (RSS) from target. We propose use particle filtering where positioning is based on predicted target's...
Salient object detection aims to locate objects that capture human attention within images. Previous approaches often pose this as a problem of image contrast analysis. In work, we model an hypergraph utilizes set hyperedges the contextual properties pixels or regions. As result, salient becomes one finding vertices and in hypergraph. The main advantage modeling is it takes into account each pixel's (or region's) affinity with its neighborhood well separation from background. Furthermore,...
In this paper, two optical image skeletonization algorithms are introduced. both algorithms, the matching between an input and a set of precalculated parallel check patterns is performed. Based on results, multistage symbolic substitution operations incorporated. For implementation, hologram-based content-addressable memory technique employed. The corresponding architecture as well character examples presented.
Sequential Monte Carlo (SMC) methods, also referred to as particle filters, have been successfully applied a variety of highly nonlinear problems such target tracking with sensor networks. In this paper, we propose the application new class SMC methods named cost-reference filters (CRPFs) mobile sensors. CRPF techniques shown be flexible and robust alternative when there is no knowledge about probability distributions noise in system. The sensors positioning during determined by predicted...
Video-based information collection has become an important research direction, and moving object tracking technique plays a key role nowadays. The classic corner algorithm doesnpsilat meet the real-time requirement, loses mostly due to occlusions, change of geometrical scale or/and some similar objects approaching object. To solve problems, new based on Kalman filter point matching estimation is proposed in paper. Combined with predicting targetpsilas location filter, extracted multi-scale...
The problem of node classification has been widely studied in a variety network-based scenarios. In this paper, we will study the more challenging scenario which some edges content-based network are labeled, and it is desirable to use information order determine labels other arbitrary edges. Furthermore, each edge associated with text content, may correspond either communication or relationship between different nodes. Such often arises context many social networks nodes, content...
The short-term photovoltaic power generation forecasting is of great significance for the system and energy management system(EMS). In this paper, model PV based on RBF neural network proposed, which forecast next 24 hours. Factors position, environment, inner performance are fully considered. A novel prediction strategy combined with mechanism used, modulations parameters executed according to online training network. Experimental results prove that proposed reduces deviation between...
Given a set of images containing objects from the same category, task image co-localization is to identify and localize each instance. This paper shows that this problem can be solved by simple but intriguing idea, is, common object detector learnt making its detection confidence scores distributed like those strongly supervised detector. More specifically, we observe given proposals extracted an contains interest, accurate should give high only small minority proposals, low most them. Thus,...