The smart vehicles construct Internet of Vehicle (IoV), which can execute various intelligent services. Although the computation capability a vehicle is limited, multi-type edge computing nodes provide heterogeneous resources for vehicular When offloading complex service to node, decision its destination should be considered according numerous factors. This paper mostly formulate as resource scheduling problem with single or multiple objective function and constraints, where some customized...

Cross-domain sentiment classification aims to address the lack of massive amounts labeled data. It demands predict polarity on a target domain utilizing classifier learned from source domain. In this paper, we investigate how efficiently apply pre-training language model BERT unsupervised adaptation. Due task and corpus, is task-agnostic, which lacks awareness can not distinguish characteristic when transferring knowledge. To tackle these problems, design post-training procedure, contains...

Video anomaly detection is commonly used in many applications, such as security surveillance, and very challenging. A majority of recent video approaches utilize deep reconstruction models, but their performance often suboptimal because insufficient error differences between normal abnormal frames practice. Meanwhile, frame prediction-based methods have shown promising performance. In this article, we propose a novel robust unsupervised method by prediction with proper design which more line...

The Internet of things (IoT) is becoming more and flexible economical with the advancement in information communication technologies. However, IoT networks will be ultra-dense explosive growth devices. Network function virtualization (NFV) emerges to provide network frameworks efficient resource management for performance networks. In NFV-enabled infrastructure, service chain (SFC) an ordered combination virtual functions (VNFs) that are related each other based on logic applications....

Federated learning is an emerging concept that trains the machine models with local distributed data sets, without sending raw to center. But, in Internet of Things (IoT) where wireless network resource constrained, key problem federated communication overhead for parameter synchronization, which wastes bandwidth, increases training time, and even impacts model accuracy. Gradient sparsification has received increasing attention, only updates significant gradients accumulates insignificant...

Space–air–ground integrated edge computing is expecting to provide pervasive computation services for Internet of Things (IoT), especially in remote areas. However, the offloading process power-limited IoT devices a challenge issue due unreliable communications an aerial environment. In this article, we propose energy-efficient space–air–ground network architecture, which choose most appropriate LEO satellites or unmanned vehicles (UAVs) task according their energy level, communication...

Chunning Du, Haifeng Sun, Jingyu Wang, Qi Qi, Jianxin Liao, Tong Xu, Ming Liu. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and 9th International Joint (EMNLP-IJCNLP). 2019.

Unmanned aerial vehicles (UAVs) can provide flexible network coverage services. UAVs be applied in a large number of scenarios, such as emergency communication and access areas without terrestrial coverage. However, are limited to relatively short range restricted energy resources. In extreme conditions disasters, there may also problem that the bandwidth is UAV cannot communicate with server amount information, so decentralized solution expected. addition, interaction between multiple...

In this paper, we propose an adaptive weighting regression (AWR) method to leverage the advantages of both detection-based and regression-based method. Hand joint coordinates are estimated as discrete integration all pixels in dense representation, guided by weight maps. This learnable aggregation process introduces supervision that allows end-to-end training brings adaptability maps, making network more accurate robust. Comprehensive exploration experiments conducted validate effectiveness...

Channel pruning can significantly accelerate and compress deep neural networks. Many channel works utilize structured sparsity regularization to zero out all the weights in some channels automatically obtain structure-sparse network training stage. However, these methods apply on each layer separately where correlations between consecutive layers are omitted. In this paper, we first combine one out-channel current corresponding in-channel next as a group, namely out-in-channel. Our proposed...

It is challenging to efficiently manage different resources in the IoT. Recently, Network function virtualization has attracted attention because of its prospect achieve efficient resource management for In NFV-enabled IoT infrastructure, a service chain (SFC) composed an ordered set virtual network functions (VNFs) that are connected based on business logic providers. However, inefficiency SFC embedding process one major problem due dynamic nature networks and abundance terminals. this...

The Internet of Vehicles (IoV) as a promising application Things (IoT) has played significant role in autonomous driving, by connecting intelligent vehicles. Autonomous driving needs to process the mass environmental sensing data coordination with surrounding vehicles, and makes an accurate judgment accordingly. Since vehicles always have limited computing resources, processing these parallel efficient task scheduling is one most important topics. Most current work focuses on formulating...

In 6G mobile systems, network slicing is an emerging technology to support services with distinct requirements by dividing a common infrastructure into multiple logical networks. However, as management method, it difficult for achieve real-time resource allocation satisfy the stringent requirement of in This paper introduces joint and routing mechanism, which combines control framework provide fine-grained, dynamic allocation. Graph Convolutional Networks (GCN)-powered Multi-Task Deep...

Compared with terrestrial networks, unmanned aerial vehicles (UAVs) have the characteristics of flexible deployment and strong adaptability, which are an important supplement to intelligent transportation systems (ITS). In this paper, we focus on multi-UAV network area coverage problem (ACP) require UAVs long-term trajectory decisions in complex scalable environment. Multi-agent deep reinforcement learning (DRL) has recently emerged as effective tool for solving problems. However, since...

Network Virtualization provides a promising tool to allow multiple heterogeneous virtual networks run on shared substrate network simultaneously. A long-standing challenge in is the Virtual Embedding (VNE) problem: how embed onto specific physical nodes and links effectively efficiently. Recent research presents several heuristic algorithms that only consider single topological attribute, which may lead decreased utilization of resources. In this paper, we introduce seven complementary...

Federated learning is an emerging concept that trains the machine models with distributed datasets, without sending raw data to center. But in edge computing enviroment where wireless network resource constrained, key problem of federated communication overhead for parameters synchronization, which wastes bandwidth, increases training time, and even impacts model accuracy. Gradient sparsification has received increasing attention, only updates significant gradients accumulates insignificant...

With the development of Internet Things (IoT) and 5G, there are ubiquitous smart devices network functions providing emerging services efficiently optimally through building many connections based on WiFi, LTE/5G, Ethernet, etc. The Multipath TCP (MPTCP) protocol that enables these to establish multiple paths for simultaneous data transmission, has been a widely used extension standard in functions. On other hand, more heavy time-varying traffic loads generated an MPTCP network, so efficient...

Depth images and point clouds are the two most commonly used data representations for depth-based 3D hand pose estimation. Benefiting from structuring of image inherent inductive biases 2D Convolutional Neural Network (CNN), image-based methods highly efficient effective. However, treating depth as a inevitably ignores nature data. Point cloud-based can better mine geometric structure these suffer disorder non-structure cloud data, which is computationally inefficient. In this paper, we...

Background Glaucoma is the leading cause of irreversible vision loss. Accurate Optic Disc (OD) and Cup (OC) segmentation beneficial for glaucoma diagnosis. In recent years, deep learning has achieved remarkable performance in OD OC segmentation. However, more challenging than due to its large shape variability cryptic boundaries that leads degradation when applying models segment OC. Moreover, are segmented independently, or pre-requirement necessary extract centered region with...

Distributed network function virtualization management and orchestration (NFV-MANO) offers a flexible way to manage orchestrate diversified services in large-scale Internet of vehicles (IoV). However, it is challenging different resources distributed NFV due the difficulties reliable message synchronization among multiple MANO systems. Recently, blockchain technology has emerged solve trust security problems for interconnections Moreover, multi-access edge computing (MEC) become prospective...

Reconstructing interacting hands from a single RGB image is very challenging task. On the one hand, severe mutual occlusion and similar local appearance between two confuse extraction of visual features, resulting in misalignment estimated hand meshes image. other there are complex spatial relationship hands, which significantly increases solution space poses difficulty network learning. In this paper, we propose decoupled iterative refinement framework to achieve pixel-alignment...