Machine reading comprehension (MRC) is an AI challenge that requires machines to determine the correct answers questions based on a given passage. MRC systems must not only answer when necessary but also tactfully abstain from answering no available according When unanswerable are involved in task, essential verification module called verifier especially required addition encoder, though latest practice modeling still mostly benefits adopting well pre-trained language models as encoder block...

Owing to the limited resources of sensor nodes, designing energy-efficient routing mechanism prolong overall network lifetime becomes one most important technologies in wireless networks (WSNs). As an active branch technology, cluster-based protocols have proven be effective topology management, energy minimization, data aggregation and so on. In this paper, we present a survey state-of-the-art techniques WSNs. We first outline clustering architecture WSNs, classify proposed approaches based...

Blind source separation (BSS) in time-frequency (TF) domain is a versatile framework to recover sources from the recorded mixture signals reverberant environment. In general, two-stage strategy one of popular BSS frameworks for underdetermined case (the number mixtures less than sources), which tough problem due mixing matrix not invertible. this paper, we propose new scheme combining density-based clustering and sparse reconstruction estimate sources, respectively. At first stage, transform...

Accurate and real-time prediction of surrounding vehicles' lane-changing intentions is a critical challenge in deploying safe efficient autonomous driving systems open-world scenarios. Existing high-performing methods remain hard to deploy due their high computational cost, long training times, excessive memory requirements. Here, we propose an intention approach based on brain-inspired Spiking Neural Networks (SNN). By leveraging the event-driven nature SNN, proposed enables us encode...

Motion estimation with variable block sizes (VBSME) is one of the most complex models in HEVC encoder. The standard supports up to 12 ranging from 4×8/8×4 64×64 for motion (ME) and compensation (MC). This feature contributes substantial coding gain compared 7 H.264/AVC at cost huge computational complexity. VBSME becomes bottleneck real time encoding. In this paper, we propose novel strategies parallel acceleration encoder based on multi-core CPU plus many-core GPU platform. Firstly, a...

Machine reading comprehension (MRC) is an AI challenge that requires machine to determine the correct answers questions based on a given passage. MRC systems must not only answer question when necessary but also distinguish no available according passage and then tactfully abstain from answering. When unanswerable are involved in task, essential verification module called verifier especially required addition encoder, though latest practice modeling still most benefits adopting well...

Meter data collection and management in smart grid has the potential for underlying security risks, e.g., low-sparsity unobservable attacks. Thus, it is crucial to investigate vulnerability of through various exposure tests associated with these Recently, much attention been paid attacks complete knowledge system matrix. In this paper, attack analysis based on a relaxed condition, i.e., an incomplete Furthermore, data-driven scheme designed demonstrate that such can be learned two-stage...

Blind source separation (BSS) is a typical unsupervised learning method that extracts latent components from their observations. In the meanwhile, convolutive BSS (CBSS) particularly challenging as observations are mixtures of well delayed versions. CBSS usually solved in frequency domain since time just instantaneous domain, which allows to recover independently each bin by running ordinary BSS, and then concatenate them form Fourier transformation signals. Because has inherent permutation...

The emerging HEVC standard supports up to 12 variable block sizes ranging from 4×8/8×4 64×64 conduct motion estimation (ME) and compensation (MC). This feature contributes considerable coding gain compared with 7 in H.264/AVC at the cost of huge computational complexity. In test model HM, ME (VBSME) may be called 425 times for mode decision procedure one CTU (Coding Tree Unit). Obviously, VBSME becomes bottleneck real time encoding. this paper, we focus on parallel realization architecture...

Underdetermined blind source separation (UBSS) is a hot topic in signal processing, which aims at recovering the signals from number of observed mixtures without knowing mixing system. Recently, expectation-maximization algorithm shows great potential UBSS. However, final results depend strongly on parameter initialization, leading to poor performance. In this paper, we propose an effective that combines tensor decomposition and nonnegative matrix factorization (NMF). proposed algorithm,...

Vehicular edge computing (VEC) has emerged as a promising paradigm to ensure the real-time task processing caused by emerging 5G or high level intelligent assisted driving applications. The tasks can be processed via services deployed at roadside units (RSUs) moving vehicles. However, dynamic topology of vehicular communication system and time-varying available resources in RSUs make challenge efficient offloading In this paper, we consider an scheme for VEC networks based on trajectory...

This work was supported in part by the National Natural Science Foundation of China under Grant 61473331, Guangdong Province 2014A030307049, Ordinary University Innovation 2015KTSCX094, Sail Plan Training High-Level Talents China, and Technology 2015B020233019, High-level Personnel Institutions Higher Learning [2013] 246.152, Scientific Research Discipline Specialty Construction Education 2013KJCX0133, 2016 Annual Technological Special Fund to foster Students Projects pdjh2016b0341,...

Generative machine reading comprehension (MRC) requires a model to generate well-formed answers. For this type of MRC, answer generation method is crucial the performance. However, generative models, which are supposed be right for task, in generally perform poorly. At same time, single-span extraction models have been proven effective extractive where constrained single span passage. Nevertheless, they suffer from generating incomplete answers or introducing redundant words when applied...

This paper proposes a new fast method for identifying the mixing matrix based on binary state mixture of Gaussian (MoG) source model. First, necessary discussion solving detection is offered under multiple dominant circumstance. Second, density presented to improve identification performance. Simulations are given demonstrate effectiveness our proposed approach.

Efficient data sharing schemes are one of the key technologies in Internet Vehicles (IoV). However, insufficient willingness vehicle users to provide makes traditional blockchain-based IoV network have low throughput. The income providers decreases when density increases on road. In this paper, we investigated a mobile scheme based consortium blockchain. detail, blockchain was used limit degree decentralization and openness, optimal revenue strategy approach between vehicles data-demand...