scispace - formally typeset
Search or ask a question
Author

Shi Dong

Bio: Shi Dong is an academic researcher from Chang'an University. The author has contributed to research in topics: Computer science & Speech coding. The author has an hindex of 15, co-authored 95 publications receiving 813 citations. Previous affiliations of Shi Dong include Southeast University & Advanced Micro Devices.


Papers
More filters
Journal ArticleDOI
TL;DR: The structural principle, the characteristics, and some kinds of classic models of deep learning, such as stacked auto encoder, deep belief network, deep Boltzmann machine, and convolutional neural network are described.

408 citations

Journal ArticleDOI
TL;DR: The state of art of the DDoS attacks in SDN and cloud computing scenarios is presented and the research works and open problems in identifying and tackling theDDoS attacks are overviewed.
Abstract: Recently, software defined networks (SDNs) and cloud computing have been widely adopted by researchers and industry. However, widespread acceptance of these novel networking paradigms has been hampered by the security threats. Advances in the processing technologies have helped attackers in increasing the attacks too, for instance, the development of Denial of Service (DoS) attacks to distributed DoS (DDoS) attacks which are seldom identified by conventional firewalls. In this paper, we present the state of art of the DDoS attacks in SDN and cloud computing scenarios. Especially, we focus on the analysis of SDN and cloud computing architecture. Besides, we also overview the research works and open problems in identifying and tackling the DDoS attacks.

125 citations

Journal ArticleDOI
TL;DR: The results of the theoretical analysis and the experimental results on datasets show that the proposed methods can better detect the DDoS attack compared with other methods.
Abstract: The Distributed Denial of Service (DDoS) attack has seriously impaired network availability for decades and still there is no effective defense mechanism against it. However, the emerging Software Defined Networking (SDN) provides a new way to reconsider the defense against DDoS attacks. In this paper, we propose two methods to detect the DDoS attack in SDN. One method adopts the degree of DDoS attack to identify the DDoS attack. The other method uses the improved K-Nearest Neighbors (KNN) algorithm based on Machine Learning (ML) to discover the DDoS attack. The results of the theoretical analysis and the experimental results on datasets show that our proposed methods can better detect the DDoS attack compared with other methods.

87 citations

Journal ArticleDOI
TL;DR: The results show that the CMSVM algorithm can reduce computation cost, improve classification accuracy and solve the imbalance problem when compared to other machine learning techniques.
Abstract: With the current massive amount of traffic that is going through the internet, internet service providers (ISPs) and networking service providers (NSPs) are looking for various ways to accurately predict the application type of flow that is going through the internet. Such prediction is critical for security and network monitoring applications as they require application type to be known in prior. Traditional ways using port-based or payload-based analysis are not sufficient anymore as many applications start using dynamic unknown port numbers, masquerading, and encryption techniques to avoid being detected. Recently, machine learning has gained significant attention in many prediction applications including traffic classification from flow features or characteristics. However, such algorithms suffer from an imbalanced data problem where some applications have fewer flow data and hence difficult to predict. In this paper, we employ network flow-level characteristics to identify the application type of traffic. Furthermore, we propose the use of an improved support vector machine (SVM) algorithm, named cost-sensitive SVM (CMSVM), to solve the imbalance problem in network traffic identification. CMSVM adopts a multi-class SVM algorithm with active learning which dynamically assigns a weight for applications. We examine the classification accuracy and performance of the CMSVM algorithm using two different datasets, namely MOORE_SET and NOC_SET datasets. Our results show that the CMSVM algorithm can reduce computation cost, improve classification accuracy and solve the imbalance problem when compared to other machine learning techniques.

67 citations

Journal ArticleDOI
TL;DR: In this paper, the effect of bio-oil on the high-temperature performance of crumb rubber modified asphalt was investigated, and the results showed that the viscosities of asphalts modified by 20mesh crumb-rubber were greater than that modified by 80mesh ones.

62 citations


Cited by
More filters
Journal Article
TL;DR: An independence criterion based on the eigen-spectrum of covariance operators in reproducing kernel Hilbert spaces (RKHSs), consisting of an empirical estimate of the Hilbert-Schmidt norm of the cross-covariance operator, or HSIC, is proposed.
Abstract: We propose an independence criterion based on the eigen-spectrum of covariance operators in reproducing kernel Hilbert spaces (RKHSs), consisting of an empirical estimate of the Hilbert-Schmidt norm of the cross-covariance operator (we term this a Hilbert-Schmidt Independence Criterion, or HSIC). This approach has several advantages, compared with previous kernel-based independence criteria. First, the empirical estimate is simpler than any other kernel dependence test, and requires no user-defined regularisation. Second, there is a clearly defined population quantity which the empirical estimate approaches in the large sample limit, with exponential convergence guaranteed between the two: this ensures that independence tests based on HSIC do not suffer from slow learning rates. Finally, we show in the context of independent component analysis (ICA) that the performance of HSIC is competitive with that of previously published kernel-based criteria, and of other recently published ICA methods.

1,134 citations

Journal ArticleDOI
TL;DR: Materials whose optical properties can be reconfigured are crucial for photonic applications such as optical memories and phase-change materials offer such utility and recent progress is reviewed.
Abstract: Materials whose optical properties can be reconfigured are crucial for photonic applications such as optical memories. Phase-change materials offer such utility and here recent progress is reviewed. Phase-change materials (PCMs) provide a unique combination of properties. On transformation from the amorphous to crystalline state, their optical properties change drastically. Short optical or electrical pulses can be utilized to switch between these states, making PCMs attractive for photonic applications. We review recent developments in PCMs and evaluate the potential for all-photonic memories. Towards this goal, the progress and existing challenges to realize waveguides with stepwise adjustable transmission are presented. Colour-rendering and nanopixel displays form another interesting application. Finally, nanophotonic applications based on plasmonic nanostructures are introduced. They provide reconfigurable, non-volatile functionality enabling manipulation and control of light. Requirements and perspectives to successfully implement PCMs in emerging areas of photonics are discussed.

872 citations

Posted Content
TL;DR: GShard enabled us to scale up multilingual neural machine translation Transformer model with Sparsely-Gated Mixture-of-Experts beyond 600 billion parameters using automatic sharding and it is demonstrated that such a giant model can efficiently be trained on 2048 TPU v3 accelerators in 4 days to achieve far superior quality for translation from 100 languages to English compared to the prior art.
Abstract: Neural network scaling has been critical for improving the model quality in many real-world machine learning applications with vast amounts of training data and compute. Although this trend of scaling is affirmed to be a sure-fire approach for better model quality, there are challenges on the path such as the computation cost, ease of programming, and efficient implementation on parallel devices. GShard is a module composed of a set of lightweight annotation APIs and an extension to the XLA compiler. It provides an elegant way to express a wide range of parallel computation patterns with minimal changes to the existing model code. GShard enabled us to scale up multilingual neural machine translation Transformer model with Sparsely-Gated Mixture-of-Experts beyond 600 billion parameters using automatic sharding. We demonstrate that such a giant model can efficiently be trained on 2048 TPU v3 accelerators in 4 days to achieve far superior quality for translation from 100 languages to English compared to the prior art.

385 citations

Journal ArticleDOI
TL;DR: In this article, a review on the techniques used to overcome/mitigate the shortcomings of conventional polymer-modified asphalt binders is provided, and a review of the effects of various types of polymers used in asphalt industry and their effects on the rheological, morphological, physical and mechanical properties of polymer modified asphalt binder are also discussed.

303 citations

Journal ArticleDOI
24 Aug 2021
TL;DR: A comprehensive review of the neural network interpretability research can be found in this paper, where a novel taxonomy organized along three dimensions: type of engagement (passive vs. active interpretation approaches), the type of explanation, and the focus (from local to global interpretability).
Abstract: Along with the great success of deep neural networks, there is also growing concern about their black-box nature. The interpretability issue affects people's trust on deep learning systems. It is also related to many ethical problems, e.g., algorithmic discrimination. Moreover, interpretability is a desired property for deep networks to become powerful tools in other research fields, e.g., drug discovery and genomics. In this survey, we conduct a comprehensive review of the neural network interpretability research. We first clarify the definition of interpretability as it has been used in many different contexts. Then we elaborate on the importance of interpretability and propose a novel taxonomy organized along three dimensions: type of engagement (passive vs. active interpretation approaches), the type of explanation, and the focus (from local to global interpretability). This taxonomy provides a meaningful 3D view of distribution of papers from the relevant literature as two of the dimensions are not simply categorical but allow ordinal subcategories. Finally, we summarize the existing interpretability evaluation methods and suggest possible research directions inspired by our new taxonomy.

237 citations