Microblog data contains rich information of real-world events with great commercial values, so microblog-oriented natural language processing (NLP) tasks have grabbed considerable attention of researchers. However, the performance of microblog-oriented Chinese Word Segmentation (CWS) based on deep neural networks (DNNs) is still not satisfying. One critical reason is that the existing microblog-oriented training corpus is inadequate to train effective weight matrices for DNNs. In this paper, we propose a novel active learning method to extend the scale of the training corpus for DNNs. However, due to a large amount of partially overlapped sentences in the microblogs, it is difficult to select samples with high annotation values from raw microblogs during the active learning procedure. To select samples with higher annotation values, parameter λ is introduced to control the number of repeatedly selected samples. Meanwhile, various strategies are adopted to measure the overall annotation values of a sample during the active learning procedure. Experiments on the benchmark datasets of NLPCC 2015 show that our λ-active learning method outperforms the baseline system and the state-of-the-art method. Besides, the results also demonstrate that the performances of the DNNs trained on the extended corpus are significantly improved.

In this letter, we propose a generalized quadrature spatial modulation technique (GQSM) which offers additional bits per channel use (bpcu) gains and a low complexity greedy detector algorithm, structured orthogonal matching pursuit (S-OMP)- GQSM, based on compressive sensing (CS) framework. Simulation results show that the bit error rate (BER) performance of the proposed greedy detector is very close to maximum likelihood (ML) and near optimal detectors based on convex programming.

By tracking the changing rate of hard decisions during every two consecutive iterations of the alternating direction method of multipliers (ADMM) penalized decoding, an efficient early termination (ET) criterion is proposed to improve the convergence rate of ADMM penalized decoder for low-density parity-check (LDPC) codes. Compared to the existing ET criterion for ADMM penalized decoding, the proposed method can reduce the average number of iterations significantly at low signal-to-noise ratios with negligible performance degradation.

The separating redundancy is an important concept in the analysis of the error-and-erasure decoding of a linear block code using a parity-check matrix of the code. In this letter, we derive new constructive upper bounds on the second separating redundancies of low-density parity-check (LDPC) codes constructed from projective and Euclidean planes over the field Fq with q even.

With the popularity of cloud computing services, outsourcing computation has entered a period of rapid development. Modular exponentiation is one of the most expensive operations in public key cryptographic systems, but the current outsourcing algorithms for modular exponentiations (MExps) with single server are inefficient or have small checkability. In this paper, we propose an efficient and fully verifiable algorithm for outsourcing multiple MExps with single untrusted server where the errors can be detected by an outsourcer with a probability of 1. The theory analysis and experimental evaluations also show that the proposed algorithm is the most efficient one compared with the previous work. Finally, we present the outsourcing schemes of digital signature algorithm (DSA) and attribute based encryption (ABE) as two applications of the proposed algorithm.

This paper presents experimental results of our proposed null-space expansion scheme for multiuser massive multiple-input multiple-output (MIMO) in time varying channels. Multiuser MIMO transmission with the proposed scheme can suppress the inter-user interference (IUI) caused by outdated channel state information (CSI). The excess degrees of freedom (DoFs) of massive MIMO is exploited to perform additional null-steering using past estimated CSI. The signal-to-interference power ratio (SIR) and spectral efficiency performances achieved by the proposed scheme that uses measured CSI is experimentally evaluated. It is confirmed that the proposed scheme shows performance superior to the conventional channel prediction scheme. In addition, IUI can be stably suppressed even in high mobility environments by further increasing the null-space dimension.

Cognitive radio (CR) is considered as the most promising solution to the so-called spectrum scarcity problem, in which channel sensing is an important problem. In this paper, the problem of determining the period of medium access control (MAC)-layer channel sensing in cognitive radio networks (CRNs) is studied. In our study, the channel state is statistically modeled as a continuous-time alternating renewal process (ARP) alternating between the OFF and ON states for the primary user (PU)'s communication activity. Based on the statistical ARP model, we analyze the CRNs with different SU MAC protocols, taking into consideration the effects of practical issues of imperfect channel sensing and non-negligible channel sensing time. Based on the analysis results, a constrained optimization problem to find the optimal sensing period is formulated and the feasibility of this problem is studied for systems with different OFF/ON channel state length distributions. Numerical results are presented to show the performance of the proposed sensing period optimization scheme. The effects of practical system parameters, including channel sensing errors and channel sensing time, on the performance and the computational complexity of the proposed sensing period optimization scheme are also investigated.

Malicious software (malware) poses various significant challenges. One is the need to retrieve plain-text messages transmitted between malware and herders through an encrypted network channel. Those messages (e.g., commands for malware) can be a useful hint to reveal their malicious activities. However, the retrieving is challenging even if the malware is executed on an analysis computer. To assist analysts in retrieving the plain-text from the memory, this paper presents FCReducer(Function Candidate Reducer), which provides a small candidate set of cryptographic functions called by malware. Given this set, an analyst checks candidates to locate cryptographic functions. If the decryption function is found, she then obtains its output as the plain-text. Although existing systems such as CipherXRay have been proposed to locate cryptographic functions, they heavily rely on fine-grained dynamic taint analysis (DTA). This makes them weak against under-tainting, which means failure of tracking data propagation. To overcome under-tainting, FCReducer conducts coarse-grained DTA and generates a typical data dependency graph of functions in which the root function accesses an encrypted message. This does not require fine-grained DTA. FCReducer then applies a community detection method such as InfoMap to the graph for detecting a community of functions that plays a role in decryption or encryption. The functions in this community are provided as candidates. With experiments using 12 samples including four malware specimens, we confirmed that FCReducer reduced, for example, 4830 functions called by Zeus malware to 0.87% as candidates. We also propose a heuristic to reduce candidates more greatly.

The rotating element electric field vector (REV) method is a classical measurement technique for phased array calibration. Compared with other calibration methods, it requires only power measurements. Thus, the REV method is more reliable for operating phased array calibration systems. However, since the phase of each element must be rotated from 0 to 2π, the conventional REV method requires a large number of measurements. Moreover, the power of composite electric field vector doesn't vary significantly because only a single element's phase is rotated. Thus, it can be easily degraded by the receiver noise. A simplified REV method combined with Hadamard group division is proposed in this paper. In the proposed method, only power measurements are required. All the array elements are divided into different groups according to the group matrix derived from the normalized Hadamard matrix. The phases of all the elements in the same group are rotated at the same time, and the composite electric field vector of this group is obtained by the simplified REV method. Hence, the relative electric fields of all elements can be obtained by a matrix equation. Compared with the conventional REV method, the proposed method can not only reduce the number of measurements but also improve the measurement accuracy under the particular range of signal to noise ratio(SNR) at the receiver, especially under low and moderate SNRs.

Ethernet passive optical network (EPON) is one of the energy-efficient access networks. Many studies have been done to reach maximum energy saving in the EPON. However, it is a trade-off between achieving maximum energy saving and guaranteeing QoS. In this paper, a predictive doze mode mechanism in an enhanced EPON architecture is proposed to achieve energy saving by using a logistic regression (LR) model. The optical line terminal (OLT) in the EPON employs an enhanced Doze Manager practicing the LR model to predict the doze periods of the optical network units (ONUs). The doze periods are estimated more accurately based on the historical high-priority traffic information, and logistic regression DBA (LR-DBA) performs dynamic bandwidth allocation accordingly. The proposed LR-DBA mechanism is compared with a scheme without energy saving (IPACT) and another scheme with energy saving (GDBA). Simulation results show that LR-DBA effectively improves the power consumption of ONUs in most cases, and the improvement can be up to 45% while it guarantees the QoS metrics, such as the high-priority traffic delay and jitter.

Multivariate Public Key Cryptosystems (MPKCs) are often touted as future-proofing against Quantum Computers. In 2009, it was shown that hardware advances do not favor just “traditional” alternatives such as ECC and RSA, but also makes MPKCs faster and keeps them competitive at 80-bit security when properly implemented. These techniques became outdated due to emergence of new instruction sets and higher requirements on security. In this paper, we review how MPKC signatures changes from 2009 including new parameters (from a newer security level at 128-bit), crypto-safe implementations, and the impact of new AVX2 and AESNI instructions. We also present new techniques on evaluating multivariate polynomials, multiplications of large finite fields by additive Fast Fourier Transforms, and constant time linear solvers.

This work presents a low-complexity lossless electrocardiogram (ECG) compression ASIC for wireless sensors. Three linear predictors aiming for different signal characteristics are provided for prediction based on a history table that records of the optimum predictors for recent samples. And unlike traditional methods using a unified encoder, the prediction error is encoded by a hybrid Golomb encoder combining Exp-Golomb and Golomb-Rice and can adaptively configure the encoding scheme according to the predictor selection. The novel adaptive prediction and encoding scheme contributes to a compression rate of 2.77 for the MIT-BIH Arrhythmia database. Implemented in 40nm CMOS process, the design takes a small gate count of 1.82K with 37.6nW power consumption under 0.9V supply voltage.

Network virtualization and slicing technologies create opportunity for infrastructure-less virtual network operators (VNOs) to enter the market anytime and provide diverse services. Multiple VNOs compete to provide the same kinds of services to end users (EUs). VNOs lease virtual resources from the infrastructure provider (InP) and sell services to the EUs by using the leased resources. The difference between the selling and leasing is the gross profit for the VNOs. A VNO that leases resources without precise knowledge of future demand, may not consume all the leased resources through service offers to EUs. Consequently, the VNO experiences loss and resources remain unused. In order to improve resource utilization and ensure that new entrant VNOs do not face losses, proper estimation of initial resource demand is important. In this paper, we propose a Bayesian game with Cournot oligopoly model to properly estimate the optimal initial resource demands for multiple entrant competing VNOs (players) with the objective of maximizing the expected profit for each VNO. The VNOs offer the same kinds of services to EUs with different qualities (player's type), which are public information. The exact service quality with which a VNO competes in the market is private information. Therefore, a VNO assumes the type of its opponent VNOs with certain probability. We derive the Bayesian Nash equilibrium (BNE) of the presented game and evaluate numerically the effect of service qualities and prices on the expected profit and market share of the VNOs.

IoT (Internet of Things) services are emerging and the bandwidth requirements for rich media communication services are increasing exponentially. We propose a virtual edge architecture comprising computation resource management layers and path bandwidth management layers for easy addition and reallocation of new service node functions. These functions are performed by the Virtualized Network Function (VNF), which accommodates terminals covering a corresponding access node to realize fast VNF migration. To increase network size for IoT traffic, VNF migration is limited to the VNF that contains the active terminals, which leads to a 20% reduction in the computation of VNF migration. Fast dynamic bandwidth allocation for dynamic bandwidth paths is realized by proposed Hierarchical Time Slot Allocation of Optical Layer 2 Switch Network, which attain the minimum calculation time of less than 1/100.

How to represent images into highly compact binary codes is a critical issue in many computer vision tasks. Existing deep hashing methods typically focus on designing loss function by using pairwise or triplet labels. However, these methods ignore the attention mechanism in the human visual system. In this letter, we propose a novel Deep Attention Residual Hashing (DARH) method, which directly learns hash codes based on a simple pointwise classification loss function. Compared to previous methods, our method does not need to generate all possible pairwise or triplet labels from the training dataset. Specifically, we develop a new type of attention layer which can learn human eye fixation and significantly improves the representation ability of hash codes. In addition, we embedded the attention layer into the residual network to simultaneously learn discriminative image features and hash codes in an end-to-end manner. Extensive experiments on standard benchmarks demonstrate that our method preserves the instance-level similarity and outperforms state-of-the-art deep hashing methods in the image retrieval application.

In this paper, we propose an automotive radar sensor compensation method improving direction of arrival (DOA) and preventing target split tracking. Amplitude and phase mismatching and mutual coupling between radar sensor arrays cause an inaccuracy problem in DOA estimation. By quantifying amplitude and phase distortion levels for each angle, we compensate the sensor distortion. Applying the proposed method to Bartlett, Capon and multiple signal classification (MUSIC) algorithms, we experimentally demonstrate the performance improvement using both experimental data from the chamber and real data obtained in actual road.

We present a GPU (graphics processing unit) accelerated stochastic algorithm implementation for simulating biochemical reaction networks using the latest NVidia architecture. To effectively utilize the massive parallelism offered by the NVidia Pascal hardware, we apply a set of performance tuning methods and guidelines such as exploiting the architecture's memory hierarchy in our algorithm implementation. Based on our experimentation results as well as comparative analysis using CPU-based implementations, we report our initial experiences on the performance of modern GPUs in the context of scientific computing.

This paper proposes a method to extract and visualize a library update history in a project. The method identifies reused library versions by comparing source code in a product with existing versions of the library so that developers can understand when their own copy of a library has been copied, modified, and updated.

With the rapid increase in demand for mobile data, mobile network operators are trying to expand wireless network capacity by deploying wireless local area network (LAN) hotspots on which they can offload their mobile traffic. However, these network-centric methods usually do not fulfill the interests of mobile users (MUs). Taking into consideration many issues, MUs should be able to decide whether to offload their traffic to a complementary wireless LAN. Our previous work studied single-flow wireless LAN offloading from a MU's perspective by considering delay-tolerance of traffic, monetary cost and energy consumption. In this paper, we study the multi-flow mobile data offloading problem from a MU's perspective in which a MU has multiple applications to download data simultaneously from remote servers, and different applications' data have different deadlines. We formulate the wireless LAN offloading problem as a finite-horizon discrete-time Markov decision process (MDP) and establish an optimal policy by a dynamic programming based algorithm. Since the time complexity of the dynamic programming based offloading algorithm is still high, we propose a low time complexity heuristic offloading algorithm with performance sacrifice. Extensive simulations are conducted to validate our proposed offloading algorithms.

This letter deals with an audio declipping problem and proposes a multiple matrix rank minimization approach. We assume that short-time audio signals satisfy the autoregressive (AR) model and formulate the declipping problem as a multiple matrix rank minimization problem. To solve this problem, an iterative algorithm is provided based on the iterative partial matrix shrinkage (IPMS) algorithm. Numerical examples show its efficiency.