With the rapid development of Internet-of-Things (IoT) and explosive advance deep learning, there is an urgent need to enable learning inference on IoT devices in Mobile Edge Computing (MEC). To address computation limitation processing complex Deep Neural Networks (DNNs), offloading proposed as a promising approach. Recently, partial developed dynamically adjust task assignment strategy different channel conditions for better performance. In this paper, we take advantage intrinsic DNN...

The increasing traffic demand has become a serious concern for cellular networks. To solve the explosion problem in vehicular network environment, there have been many efforts to offload from links Roadside Units (RSUs). Compared with cost of downloading link, through RSUs is considered practically free. In most cases, we wait one or several download entire data, which causing huge delays. However, people can always data network. reality, are sensitive delay but would like pay little money...

Urban tracking plays a vital role for people's urban life in intelligent transportation systems, e.g., public safety, case investigation, finding missing items, etc. However, the current methods consume large amount of communication and computing resources since they mainly offload all related sensing data, i.e., videos, generated by widely deployed cameras to cloud where data are stored, processed, analyzed. In this paper, we propose graph optimized offloading algorithm leveraging crowd-AI...

With the progress of mobile devices and successful forms using wisdom crowds, spatial crowdsourcing has attracted much attention from research community. The idea is recruiting a set available crowds to finish tasks located in locations, e.g., landmarks, by their handheld devices. This paper addresses worker recruitment problem under coverage workload-balancing requirements. constraint means that any location should be visited at least one recruited workers satisfy Quality-of-Service...

Due to the additive property of most machine learning objective functions, training can be distributed multiple machines. Distributed is an efficient way deal with rapid growth data volume at cost extra inter-machine communication. One common implementation parameter server system which contains two types nodes: worker nodes, are used for calculating updates, and maintaining parameters. We observe that inefficient communication between workers servers may slow down system. Therefore, we...

A mobile ad hoc network (MANET) is a self-configuring infrastructure-less network. Taking advantage of spontaneous and behavior, MANET can be integrated with satellite to provide world-wide communication for emergency disaster relieve services also cellular data offloading. To achieve different purposes, architecture system, protocols mechanisms are designed. For services, ubiquitous robust communications paramount importance. offloading emphasis amount offloaded data, limited storage energy...

Computation offloading is proposed to solve one obstacle of enabling high-accurate and real-time deep inference in resource-constrained Internet Things (IoT) devices. Cooperative recently further trade-off the introduced communication latency computation offloading, which partitions a Deep Neural Network (DNN) model into two parts utilizes IoT end device server process DNN cooperatively. We observe important but ignored fact all previous works: processing cbe conducted simultaneously...

In opportunistic mobile networks, existing schemes rely on the assumption that data can be entirely transmitted at each contact. However, in an network, transmission probability exponentially decreases as size increases. That is, contact duration might insufficient to deliver large data. Therefore, it is reasonable partition original into small chunks and chunk forwarded through path. The objective of this paper find optimal strategy where delivery ratio maximized under a given deadline....

In cellular traffic offloading through opportunistic mobile networks, existing schemes rely on the assumption that data can be entirely transmitted at each contact. However, transmission probability exponentially decreases as size increases. That is, contact duration in might insufficient for delivering large data. The objective of this paper is to find an optimal scheme partitioning so delivery latency minimized. There a trade- off partitioning. Each small chunk path has higher than...

In this paper, a general data and acknowledgement dissemination mechanism is proposed in mobile social networks (MSNs). Most existing schemes MSNs only consider transmission. However, receiving has many potential applications (e.g., trade incentive mechanism). Challenging problems thus arise due to type of mixed messages (i.e., acknowledgement) problem. The buffer constraint time for make problem even harder handle practical scenario. order maximize the research objective low delay high...

This paper addresses the coverage and workload-balancing requirements of worker recruiting in spatial crowdsourcing. That is, recruited workers should be able to visit all crowdsourcing locations satisfy a certain quality, e.g., traffic monitoring or climate forecast. In addition, each operation has cost, data energy consumption, location might have budget for visited workers. The objective this is find algorithm, which ensures requirement minimizes maximal cost any location. We gradually...

Data partitioning is the most fundamental procedure before parallelizing complex analysis on very big graphs. As a classical NP-complete problem, graph usually employs offline or online/streaming heuristics to find approximately optimal solutions. However, they are either heavyweight in space and time overheads suboptimal quality measured by workload balance number of cutting edges across partitions, both which cannot scale well with ever-growing demands quickly analyzing This paper thereby...

The vigorous development of IoT technology has spawned a series applications that are delay-sensitive or resource-intensive. Mobile edge computing is an emerging paradigm provides services between end devices and traditional cloud data centers to users. However, with the continuously increasing investment demands, it nontrivial maintain higher quality-of-service (QoS) under erratic activities mobile In this paper, we investigate service provisioning updating problem multiple-users scenario...

In the last decade, Convolutional Neural Network with a multi-layer architecture has advanced rapidly. However, training its complex network is very space-consuming, since lot of intermediate data are preserved across layers, especially when processing high-dimension inputs big batch size. That poses great challenges to limited memory capacity current accelerators (e.g., GPUs). Existing efforts mitigate such bottleneck by external auxiliary solutions additional hardware costs, and internal...

The topic-based mobile publish/subscribe (pub/sub) system has shown the potential applications in many scenarios, e.g., product coupon distribution. In this paper, we focus on budget-constrained data dissemination services with a pre-determinated total amount of copy. A user may subscribe under different topics, but receiving copy any topic is enough. This mutually exclusive delivery requirement scenarios. light amounts and popularity levels each topic, deciding which should be forwarded to...

In Mobile Social Networks (MSNs), the single-path routing might not have enough of a chance to transmit content destination (i.e., low network throughput), due limited contact opportunities. Meanwhile, multiple-path improves throughput at cost higher system resource consumption (e.g., energy and storage). Therefore, there exists trade-off between consumption. Moreover, we should consider user features in MSNs, i.e., some nodes would like help other with same social neighbors, classmates)...

To improve the user experience in multi-party online applications, i.e., low gaming lag gaming, maximum latency between any pair of users should be minimized. Considering multiple preferences users, which group a can join, we address User Latency Minimization (ULM) problem by performing an optimal matchmaking. This paper proves that ULM is NP-hard if have more than two preferences. We prove has sub-modular property and apply classic greedy algorithm with approximation ratio 1+ n, where n...

With the progress of mobile devices and successful usage wisdom crowds, spatial crowdsourcing has attracted much attention from research community. This paper addresses efficient worker recruitment problem under task coverage constraint. The efficiency is measured by total quality collected a set workers corresponding cost, e.g., proportional to overall trajectory length workers. Specifically, we consider two different scenarios, 1-D line topology general 2-D topology, in which may have...

Most existing dissemination schemes in Mobile Social Networks (MSNs) only consider the data dissemination. However, there are two types of messages: and control message (i.e., acknowledgment) MSNs, receiving acknowledgment is very important many applications (e.g., mobile trade incentive mechanism). In order to maximize desired delivery ratio, we have identify priority each network during limited contact opportunity. Therefore, propose a generic priority-based compare-split routing scheme,...

The capability's of our mobile devices, e.g., smartphones, allows for multiple connectivity interfaces and overall a better experience sending receiving data. limiting factor that slows down the process is latency through cellular WiFi connections within different environments. By exploring devices in connection scenarios, it opens up tons opportunity's to find more cost-efficient ways connect networks, download applications, use best possible interface given scenario. In this paper, we...

WaterMobile and WaterTalk are two activities that were developed for K-12 students educators to assist in workforce development the water/wastewater utilities. Utilities involved many types of water related relate treatment, understanding forces water, measurement quality parameters much more. These specifically bring forth STEM careers relevant