A5100447258- Network Security and Intrusion Detection
- Advanced Computational Techniques and Applications
- Neural Networks and Applications
- Advanced Graph Neural Networks
- Advanced Malware Detection Techniques
- Advanced Algorithms and Applications
- Internet Traffic Analysis and Secure E-voting
- Anomaly Detection Techniques and Applications
- Adversarial Robustness in Machine Learning
- Topic Modeling
- Recommender Systems and Techniques
- Service-Oriented Architecture and Web Services
- Software System Performance and Reliability
- Quantum Information and Cryptography
- Advanced Decision-Making Techniques
- Software-Defined Networks and 5G
- Cloud Computing and Resource Management
- Rough Sets and Fuzzy Logic
- Caching and Content Delivery
- Quantum Computing Algorithms and Architecture
- Fuzzy Logic and Control Systems
- Power Systems and Technologies
- Data Mining Algorithms and Applications
- Technology and Security Systems
- Radiomics and Machine Learning in Medical Imaging
Zhejiang University
2012-2024
Peking University
2005-2024
Meteorological Bureau of Shenzhen Municipality
2024
Capital University
2023
Sichuan University
2023
Soochow University
2017-2023
Children's Hospital of Suzhou University
2023
Second Affiliated Hospital of Xi'an Jiaotong University
2023
Liaoning Technical University
2023
Sichuan Agricultural University
2023
We present SWAN, a system that boosts the utilization of inter-datacenter networks by centrally controlling when and how much traffic each service sends frequently re-configuring network's data plane to match current demand. But done simplistically, these re-configurations can also cause severe, transient congestion because different switches may apply updates at times. develop novel technique leverages small amount scratch capacity on links in provably congestion-free manner, without making...
While many public cloud providers offer pay-as-you-go computing, their varying approaches to infrastructure, virtualization, and software services lead a problem of plenty. To help customers pick that fits needs, we develop CloudCmp, systematic comparator the performance cost providers. CloudCmp measures elastic persistent storage, networking offered by along metrics directly reflect impact on customer applications. strives ensure fairness, representativeness, compliance these measurements...
Two novel channel-assignment strategies are proposed: the locally optimized dynamic assignment (LODA) strategy and borrowing with directional channel-locking (BDCL) strategy. Their performance is compared fixed-assignment (FA) (currently used on certain systems) channel ordering (BCO) (the that has given lowest blocking probability in previous research). Computer simulations a 49-cell network for both uniform nonuniform traffic showed average call-blocking of BDCL always lowest. The LODA...
We present SWAN, a system that boosts the utilization of inter-datacenter networks by centrally controlling when and how much traffic each service sends frequently re-configuring network's data plane to match current demand. But done simplistically, these re-configurations can also cause severe, transient congestion because different switches may apply updates at times. develop novel technique leverages small amount scratch capacity on links in provably congestion-free manner, without making...
Breast cancer is one of the deadliest cancers that cause women death globally. Ultrasound imaging commonly used diagnostic tools for detection and classification breast abnormalities. In past decades, computer-aided diagnosis (CAD) systems have been developed to improve accuracy made by radiologists. particular, automatic ultrasound (BUS) image segmentation a critical step using CAD. However, accurate tumor still challenge as result various artifacts. This paper novel framework based on deep...
Inter-data center wide area networks (inter-DC WANs) carry a significant amount of data transfers that require to be completed within certain time periods, or deadlines. However, very little work has been done guarantee such The crux is the current inter-DC WAN lacks an interface for users specify their transfer deadlines and mechanism provider ensure completion while maintaining high utilization. In this paper, we address problem by introducing deadline-based network abstraction (DNA) WANs....
An algorithm for allocating nominal channels according to traffic distribution is designed. The attempts minimize the average blocking probability as are allocated one at a time. Simulation results show that system's traffic-carrying capacity can be increased by about 10% use of this algorithm, and gain additional improvement obtained from channel-borrowing strategies. If effect shadow considered in assignment channels, only very small increase observed.< <ETX...
Today, large-scale web services run on complex systems, spanning multiple data centers and content distribution networks, with performance depending diverse factors in end infrastructure servers. Web service providers have many options for improving performance, varying greatly feasibility, cost benefit, but few tools to predict the impact of these options.A key challenge is precisely capture object dependencies, as are essential predicting an accurate scalable manner. In this paper, we...
Inter-datacenter wide area networks (inter-DC WAN) carry a significant amount of data transfers that require to be completed within certain time periods, or deadlines. However, very little work has been done guarantee such The crux is the current inter-DC WAN lacks an interface for users specify their transfer deadlines and mechanism provider ensure completion while maintaining high utilization.
It was recently shown that the gut microbiota of both depression patients and model animals is significantly altered, suggesting microbes are closely related to depression. Here, we investigated effects Sophora alopecuroides L.-derived alkaloids on mice with depression-like behaviors. We first established a mouse via chronic unpredictable mild stress (CUMS) detected changes in behaviors depression-related indicators. Simultaneously, 16S rRNA sequencing performed investigate changes. improved...
A piecewise second order approximation scheme is proposed for computing the sigmoid function. The provides high performance with low implementation cost; thus, it suitable hardwired cost effective neural emulators. It shown that an of generator outperforms, in both precision and speed, existing schemes using a bit serial pipelined implementation. requires one multiplication, no look-up table addition. has been estimated output generated maximum computation delay 21 machine cycles...
Logs are imperative in the management process of networks and services. However, manually identifying classifying anomalous logs is time-consuming, error-prone, labor-intensive. Additionally, rule-based approaches cannot tackle challenges underlying log identification classification resulting from new types partial labels. We propose LogClass, a framework to automatically robustly identify classify for network service based on <italic xmlns:mml="http://www.w3.org/1998/Math/MathML"...
As the machine learning-related technology has made great progress in recent years, deep neural networks are widely used many scenarios, including security-critical ones, which may incura loss when DNN is compromised. Starting from introducing several commonly bit-flip methods, this paper concentrates on bit-flips attacks aiming and corresponding defense methods. We analyze threat models, methods design, effect of attack detail, drawing some helpful conclusions about improving robustness...
Intrusion detection occupies a decision position in solving the network security problems. Support Vector Machines (SVMs) are one of widely used intrusion techniques. However, commonly two-class SVM algorithms facing difficulties constructing training dataset. That is because many real application scenarios, normal connection records easy to be obtained, but attack not so. We propose an anomaly model based on One-class detect intrusions. The one-class adopts only as But after being trained,...
Heterogeneous information networks (HINs) is a general representation of many real world applications. The difference between HIN and traditional homogeneous graphs that the nodes edges in are with types. Then applications, we need to consider types make approach more semantically meaningful. For applications annotation expensive, on natural way semi-supervised learning over HIN. In this paper, present algorithm constrained by HINs. We first decompose original into several meaningful...
Knowledge graph inference has been studied extensively due to its wide applications. It addressed by two lines of research, i.e., the more traditional logical rule reasoning and recent knowledge embedding (KGE). Several attempts have made combine KGE rules for better inference. Unfortunately, they either simply treat as additional constraints into loss or use probabilistic model approximate exact (i.e., MAX-SAT). Even worse, both approaches need sample ground tackle scalability issue, total...
Training the generative models with minimal corpus is one of critical challenges for building open-domain dialogue systems. Existing methods tend to use meta-learning framework which pre-trains parameters on all non-target tasks then fine-tunes target task. However, fine-tuning distinguishes from parameter perspective but ignores model-structure perspective, resulting in similar different tasks. In this paper, we propose an algorithm that can customize a unique model each task few-shot...