Analysis of Internet Things (IoT) sensor data is a key for achieving city smartness. In this paper multitier fog computing model with large-scale analytics service proposed smart cities applications. The consisted ad-hoc fogs and dedicated opportunistic resources, respectively. new clear functional modules able to mitigate the potential problems infrastructure slow response in cloud computing. We run benchmark experiments over formed by Rapsberry Pi computers distributed engine measure...

Acquiring sufficient ground-truth supervision to train deep vi- sual models has been a bottleneck over the years due data-hungry nature of learning. This is exacerbated in some structured prediction tasks, such as semantic segmen- tation, which requires pixel-level annotations. work ad- dresses weakly supervised segmentation (WSSS), with goal bridging gap between image-level anno- tations and segmentation. We formulate WSSS novel group-wise learning task that explicitly se- mantic...

Sketches, a type of probabilistic algorithms, have been widely accepted as the approximate summary data streams. Compressing sketches is best choice in distributed streams to reduce communication overhead. The ideal compression algorithm should meet following three requirements: high efficiency procedure, support direct query without decompression, and accuracy compressed sketches. However, no prior work can these requirements at same time. Especially, poor after using existing methods. In...

In this paper, we introduce an asynchronous decentralized accelerated stochastic gradient descent type of algorithm for optimization. Considering communication and synchronization costs are the major bottlenecks optimization, attempt to reduce these from algorithmic design aspect, in particular, able number agents involved one round update via randomization. Our contribution is develop a class randomized algorithms solving general convex composite problems. We establish <i...

Micro-blogging service has been developing and evolving rapidly in China which led to a significant rise social spamming attacks.However, little is known about these spammers.Thus, this paper, we presented an observation on spammers Sina Weibo, the biggest micro-blogging community China.Specifically, used program-controlled profiles monitor, track record behaviors.We gave detailed description of experiment settings then analyzed data collected by profiles.We found that Weibo can be...

One fundamental problem in constrained decentralized multiagent optimization is the trade-off between gradient/sampling complexity and communication complexity. In this paper, we propose new algorithms whose gradient sampling complexities are graph topology invariant, while their remain optimal. Specifically, for convex smooth deterministic problems, a primal-dual sliding (PDS) algorithm that able to compute an -solution with complexity, where smoothness parameter of objective function...

We present a new class of decentralized first-order methods for nonsmooth and stochastic optimization problems defined over multiagent networks. Considering that communication is major bottleneck in optimization, our main goal this paper to develop algorithmic frameworks which can significantly reduce the number inter-node communications. first propose primal-dual method find an $\epsilon$-solution both terms functional optimality gap feasibility residual $O(1/\epsilon)$ rounds when...

In this work, we introduce an asynchronous decentralized accelerated stochastic gradient descent type of method for optimization, considering communication and synchronization are the major bottlenecks. We establish $\mathcal{O}(1/\epsilon)$ (resp., $\mathcal{O}(1/\sqrt{\epsilon})$) complexity $\mathcal{O}(1/\epsilon^2)$ $\mathcal{O}(1/\epsilon)$) sampling solving general convex strongly convex) problems.

Distributionally robust optimization (DRO) is a powerful framework for training models against data distribution shifts. This paper focuses on constrained DRO, which has an explicit characterization of the robustness level. Existing studies DRO mostly focus convex loss function, and exclude practical challenging case with non-convex e.g., neural network. develops stochastic algorithm its performance analysis DRO. The computational complexity our at each iteration independent overall dataset...

Due to the advent of electronic commerce, need for flexibility, adaptability and rapid response time information services system has become increasingly important. In order cope with continuously changing conditions service provision utilization, Faded Information Field (FIF) been proposed. FIF is a distributed architecture that sustained by push/pull mobile agents bring high-assurance through users' preference-oriented prevision on supply-demand basis. this paper we propose technology...

GraphChi is the first reported disk-based graph engine that can handle billion-scale graphs on a single PC efficiently. able to execute several advanced data mining, mining and machine learning algorithms very large graphs. With novel technique of parallel sliding windows (PSW) load subgraph from disk memory for vertices edges updating, it achieve processing performance close even better than those mainstream distributed engines. mentioned its not effectively utilized with dataset, which...

Current information service systems are characterized by continuous changes of services from business companies promoting their special offers on-line, and unpredictable users' demand. In such system, providers (SPs) require timely update to inform about new products, at the same time, users a access services. This fosters an urgent need for design high-assurance that is, system can meet both providers' heterogeneous requirements. Faded Information Field (FIF), sustained push/pull mobile...

Color histogram is widely used in the appearance modeling. However lack of spatial information makes features less distinctive. In this paper, we propose to detect local region relationship based on segmentation color space. After efficiently computing region's information, a novel topological histograms represent context. Through experimental observation, find model make object's feature more discriminative.

In order to meet the heterogeneous requirements from service providers and users simultaneously, there is an urgent need for a new information system. After studying characteristics of users' accessing, that popularity utilization generally exists in system focused. This paper introduces rating oriented distributed sustained by push/pull mobile agents provision utilization. Based on this environment, autonomous allocation technology proposed. However, when demand changes, congestion might...