Network embedding is an important method to learn low-dimensional representations of vertexes in networks, aiming capture and preserve the network structure. Almost all existing methods adopt shallow models. However, since underlying structure complex, models cannot highly non-linear structure, resulting sub-optimal representations. Therefore, how find a that able effectively global local open yet problem. To solve this problem, paper we propose Structural Deep Embedding method, namely SDNE....

Graph embedding algorithms embed a graph into vector space where the structure and inherent properties of are preserved. The existing methods cannot preserve asymmetric transitivity well, which is critical property directed graphs. Asymmetric depicts correlation among edges, that is, if there path from u to v, then likely edge v. can help in capturing structures graphs recovering partially observed To tackle this challenge, we propose idea preserving by approximating high-order proximity...

Deep learning has been shown to be successful in a number of domains, ranging from acoustics, images, natural language processing. However, applying deep the ubiquitous graph data is non-trivial because unique characteristics graphs. Recently, substantial research efforts have devoted methods graphs, resulting beneficial advances analysis techniques. In this survey, we comprehensively review different types on We divide existing into five categories based their model architectures and...

Network embedding assigns nodes in a network to low-dimensional representations and effectively preserves the structure. Recently, significant amount of progresses have been made toward this emerging analysis paradigm. In survey, we focus on categorizing then reviewing current development methods, point out its future research directions. We first summarize motivation embedding. discuss classical graph algorithms their relationship with Afterwards primarily, provide comprehensive overview...

Network embedding, aiming to learn the low-dimensional representations of nodes in networks, is paramount importance many real applications. One basic requirement network embedding preserve structure and inherent properties networks. While previous methods primarily microscopic structure, such as first- second-order proximities nodes, mesoscopic community which one most prominent feature largely ignored. In this paper, we propose a novel Modularized Nonnegative Matrix Factorization (M-NMF)...

This article introduces the principal concepts of multimedia cloud computing and presents a novel framework. We address from multimedia-aware (media cloud) cloud-aware (cloud media) perspectives. First, we present cloud, which addresses how can perform distributed processing storage provide quality service (QoS) provisioning for services. To achieve high QoS services, propose media-edge (MEC) architecture, in storage, central unit (CPU), graphics (GPU) clusters are presented at edge to...

Curriculum learning (CL) is a training strategy that trains machine model from easier data to harder data, which imitates the meaningful order in human curricula. As an easy-to-use plug-in, CL has demonstrated its power improving generalization capacity and convergence rate of various models wide range scenarios such as computer vision natural language processing etc. In this survey article, we comprehensively review aspects including motivations, definitions, theories, applications. We...

Graph Convolutional Networks (GCNs) are an emerging type of neural network model on graphs which have achieved state-of-the-art performance in the task node classification. However, recent studies show that GCNs vulnerable to adversarial attacks, i.e. small deliberate perturbations graph structures and attributes, poses great challenges for applying real world applications. How enhance robustness remains a critical open problem. To address this problem, we propose Robust GCN (RGCN), novel...

Depression is a major contributor to the overall global burden of diseases. Traditionally, doctors diagnose depressed people face via referring clinical depression criteria. However, more than 70% patients would not consult at early stages depression, which leads further deterioration their conditions. Meanwhile, are increasingly relying on social media disclose emotions and sharing daily lives, thus have successfully been leveraged for helping detect physical mental Inspired by these, our...

We have witnessed the tremendous growth of videos over Internet, where most these are typically paired with abundant sentence descriptions, such as video titles, captions and comments. Therefore, it has been increasingly crucial to associate specific segments corresponding informative text for a deeper understanding content. This motivates us explore an overlooked problem in research community — temporal localization video, which aims automatically determine start end points given within...

Exponential growth of information generated by online social networks demands effective recommender systems to give useful results. Traditional techniques become unqualified because they ignore relation data; existing recommendation approaches consider network structure, but context has not been fully considered. It is significant and challenging fuse contextual factors which are derived from users' motivation behaviors into recommendation. In this paper, we investigate on the basis...

Facial makeup transfer aims to translate the style from a given reference face image another non-makeup one while preserving identity. Such an instance-level problem is more challenging than conventional domain-level tasks, especially when paired data unavailable. Makeup also different global styles (e.g., paintings) in that it consists of several local styles/cosmetics, including eye shadow, lipstick, foundation, and so on. Extracting transferring such delicate information infeasible for...

Network embedding has received increasing research attention in recent years. The existing methods show that the high-order proximity plays a key role capturing underlying structure of network. However, two fundamental problems preserving remain unsolved. First, all can only preserve fixed-order proximities, despite proximities different orders are often desired for distinct networks and target applications. Second, given certain order proximity, cannot guarantee accuracy efficiency...

Temporal sentence grounding in videos aims to localize one target video segment, which semantically corresponds a given sentence. Unlike previous methods mainly focusing on matching semantics between the and different segments, this paper, we propose novel semantic conditioned dynamic modulation (SCDM) mechanism, leverages modulate temporal convolution operations for better correlating composing sentence-relevant contents over time. The proposed SCDM also performs dynamically with respect...

Deep learning has been shown to be successful in a number of domains, ranging from acoustics, images, natural language processing. However, applying deep the ubiquitous graph data is non-trivial because unique characteristics graphs. Recently, substantial research efforts have devoted methods graphs, resulting beneficial advances analysis techniques. In this survey, we comprehensively review different types on We divide existing into five categories based their model architectures and...

Network embedding aims to preserve vertex similarity in an space. Existing approaches usually define the by direct links or common neighborhoods between nodes, i.e. structural equivalence. However, vertexes which reside different parts of network may have similar roles positions, regular equivalence, is largely ignored literature embedding. Regular equivalence defined a recursive way that two regularly equivalent neighbors are also equivalent. Accordingly, we propose new approach named Deep...

To learn a sequential recommender, the existing methods typically adopt sequence-to-item (seq2item) training strategy, which supervises sequence model with user's next behavior as label and past behaviors input. The seq2item however, is myopic usually produces non-diverse recommendation lists. In this paper, we study problem of mining extra signals for supervision by looking at longer-term future. There exist two challenges: i) reconstructing future containing many exponentially harder than...

Network embedding has recently attracted lots of attentions in data mining. Existing network methods mainly focus on networks with pairwise relationships. In real world, however, the relationships among points could go beyond pairwise, i.e., three or more objects are involved each relationship represented by a hyperedge, thus forming hyper-networks. These hyper-networks pose great challenges to existing when hyperedges indecomposable, that is say, any subset nodes hyperedge cannot form...

Disentangled Representation Learning (DRL) aims to learn a model capable of identifying and disentangling the underlying factors hidden in observable data representation form. The process separating variation into variables with semantic meaning benefits learning explainable representations data, which imitates meaningful understanding humans when observing an object or relation. As general strategy, DRL has demonstrated its power improving explainability, controlability, robustness, as well...

In this paper, the network planning problem in wireless ad hoc networks is formulated as of allocating physical and medium access layer resources or supplies to minimize a cost function, while fulfilling certain end-to-end communication demands, which are given collection multicast sessions with desired transmission rates. We propose an iterative cross-layer optimization, alternates between: 1) jointly optimizing timesharing sum max flows assignment 2) updating operational states layer....

Roaming across heterogeneous wireless networks such as wide area network (WWAN) and local (WLAN) poses considerable challenges, it is usually difficult to maintain the existing connections guarantee necessary quality-of-service. This paper proposes a novel seamless proactive end-to-end mobility management system, which can based on principle by incorporating an intelligent status detection mechanism. The proposed system consists of two components, connection manager (CM) virtual connectivity...

Exponential growth of information generated by online social networks demands effective and scalable recommender systems to give useful results. Traditional techniques become unqualified because they ignore relation data; existing recommendation approaches consider network structure, but contextual has not been fully considered. It is significant challenging fuse factors which are derived from users' motivation behaviors into recommendation. In this paper, we investigate the problem on basis...

Today's Internet has witnessed an increase in the popularity of mobile video streaming, which is expected to exceed 3/4 global data traffic by 2019. To satisfy considerable amount requests, service providers have been pushing their content delivery infrastructure edge networks-from regional network (CDN) servers peer CDN (e.g., smartrouters users' homes)-to cache and serve users with storage resources nearby. Among caching paradigms, Wi-Fi access point cellular base station become two...

Cascades are ubiquitous in various network environments such as epidemic networks, traffic water distribution networks and social networks. The outbreaks of cascades will often bring bad or even devastating effects. How to accurately predict the cascading early stage is paramount importance for people avoid these Although there have been some pioneering works on detection, how predict, rather than detect, still an open problem. In this paper, we attempt harnessing historical cascade data,...