arXiv subjects

Li Xu

Publications and source records attributed to Li Xu.

124 records · Page 7Linked to original sources

arXiv2015

ROSA: Robust sparse adaptive channel estimation in the presence of impulsive noises

Based on the assumption of Gaussian noise model, conventional adaptive filtering algorithms for reconstruction sparse channels were proposed to take advantage of channel sparsity due to the fact that broadband wireless channels usually have the sparse nature. However, state-of-the-art algorithms are vulnerable to deteriorate under the assumption of non-Gaussian noise models (e.g., impulsive noise) which often exist in many advanced communications systems. In this paper, we study the problem of RObust Sparse Adaptive channel estimation (ROSA) in the environment of impulsive noises using variable step-size affine projection sign algorithm (VSS-APSA). Specifically, standard VSS-APSA algorithm is briefly reviewed and three sparse VSS-APSA algorithms are proposed to take advantage of channel sparsity with different sparse constraints. To fairly evaluate the performance of these proposed algorithms, alpha-stable noise is considered to approximately model the realistic impulsive noise environments. Simulation results show that the proposed algorithms can achieve better performance than standard VSS-APSA algorithm in different impulsive environments.

cs.IT↗

arXiv2015

Robust Adaptive Sparse Channel Estimation in the Presence of Impulsive Noises

Broadband wireless channels usually have the sparse nature. Based on the assumption of Gaussian noise model, adaptive filtering algorithms for reconstruction sparse channels were proposed to take advantage of channel sparsity. However, impulsive noises are often existed in many advance broadband communications systems. These conventional algorithms are vulnerable to deteriorate due to interference of impulsive noise. In this paper, sign least mean square algorithm (SLMS) based robust sparse adaptive filtering algorithms are proposed for estimating channels as well as for mitigating impulsive noise. By using different sparsity-inducing penalty functions, i.e., zero-attracting (ZA), reweighted ZA (RZA), reweighted L1-norm (RL1) and Lp-norm (LP), the proposed SLMS algorithms are termed as SLMS-ZA, SLMS-RZA, LSMS-RL1 and SLMS-LP. Simulation results are given to validate the proposed algorithms.

cs.IT↗

arXiv2015

Improved Adaptive Sparse Channel Estimation Using Re-Weighted L1-norm Normalized Least Mean Fourth Algorithm

In next-generation wireless communications systems, accurate sparse channel estimation (SCE) is required for coherent detection. This paper studies SCE in terms of adaptive filtering theory, which is often termed as adaptive channel estimation (ACE). Theoretically, estimation accuracy could be improved by either exploiting sparsity or adopting suitable error criterion. It motivates us to develop effective adaptive sparse channel estimation (ASCE) methods to improve estimation performance. In our previous research, two ASCE methods have been proposed by combining forth-order error criterion based normalized least mean fourth (NLMF) and L1-norm penalized functions, i.e., zero-attracting NLMF (ZA-NLMF) algorithm and reweighted ZA-NLMF (RZA-NLMF) algorithm. Motivated by compressive sensing theory, an improved ASCE method is proposed by using reweighted L1-norm NLMF (RL1-NLMF) algorithm where RL1 can exploit more sparsity information than ZA and RZA. Specifically, we construct the cost function of RL1-NLMF and hereafter derive its update equation. In addition, intuitive figure is also given to verify that RL1 is more efficient than conventional two sparsity constraints. Finally, simulation results are provided to confirm this study.

cs.IT↗

arXiv2015

On Vectorization of Deep Convolutional Neural Networks for Vision Tasks

We recently have witnessed many ground-breaking results in machine learning and computer vision, generated by using deep convolutional neural networks (CNN). While the success mainly stems from the large volume of training data and the deep network architectures, the vector processing hardware (e.g. GPU) undisputedly plays a vital role in modern CNN implementations to support massive computation. Though much attention was paid in the extent literature to understand the algorithmic side of deep CNN, little research was dedicated to the vectorization for scaling up CNNs. In this paper, we studied the vectorization process of key building blocks in deep CNNs, in order to better understand and facilitate parallel implementation. Key steps in training and testing deep CNNs are abstracted as matrix and vector operators, upon which parallelism can be easily achieved. We developed and compared six implementations with various degrees of vectorization with which we illustrated the impact of vectorization on the speed of model training and testing. Besides, a unified CNN framework for both high-level and low-level vision tasks is provided, along with a vectorized Matlab implementation with state-of-the-art speed performance.

cs.CV↗

arXiv2015

Iterative-Promoting Variable Step Size Least Mean Square Algorithm for Accelerating Adaptive Channel Estimation

Invariable step size based least-mean-square error (ISS-LMS) was considered as a very simple adaptive filtering algorithm and hence it has been widely utilized in many applications, such as adaptive channel estimation. It is well known that the convergence speed of ISS-LMS is fixed by the initial step-size. In the channel estimation scenarios, it is very hard to make tradeoff between convergence speed and estimation performance. In this paper, we propose an iterative-promoting variable step size based least-mean-square error (VSS-LMS) algorithm to control the convergence speed as well as to improve the estimation performance. Simulation results show that the proposed algorithm can achieve better estimation performance than previous ISS-LMS while without sacrificing convergence speed.

cs.IT↗

arXiv2014

An Evasion and Counter-Evasion Study in Malicious Websites Detection

Malicious websites are a major cyber attack vector, and effective detection of them is an important cyber defense task. The main defense paradigm in this regard is that the defender uses some kind of machine learning algorithms to train a detection model, which is then used to classify websites in question. Unlike other settings, the following issue is inherent to the problem of malicious websites detection: the attacker essentially has access to the same data that the defender uses to train its detection models. This 'symmetry' can be exploited by the attacker, at least in principle, to evade the defender's detection models. In this paper, we present a framework for characterizing the evasion and counter-evasion interactions between the attacker and the defender, where the attacker attempts to evade the defender's detection models by taking advantage of this symmetry. Within this framework, we show that an adaptive attacker can make malicious websites evade powerful detection models, but proactive training can be an effective counter-evasion defense mechanism. The framework is geared toward the popular detection model of decision tree, but can be adapted to accommodate other classifiers.

cs.CR↗

arXiv2014

Extra Gain:Improved Sparse Channel Estimation Using Reweighted l_1-norm Penalized LMS/F Algorithm

The channel estimation is one of important techniques to ensure reliable broadband signal transmission. Broadband channels are often modeled as a sparse channel. Comparing with traditional dense-assumption based linear channel estimation methods, e.g., least mean square/fourth (LMS/F) algorithm, exploiting sparse structure information can get extra performance gain. By introducing l_1-norm penalty, two sparse LMS/F algorithms, (zero-attracting LMSF, ZA-LMS/F and reweighted ZA-LMSF, RZA-LMSF), have been proposed [1]. Motivated by existing reweighted l_1-norm (RL1) sparse algorithm in compressive sensing [2], we propose an improved channel estimation method using RL1 sparse penalized LMS/F (RL1-LMS/F) algorithm to exploit more efficient sparse structure information. First, updating equation of RL1-LMS/F is derived. Second, we compare their sparse penalize strength via figure example. Finally, computer simulation results are given to validate the superiority of proposed method over than conventional two methods.

cs.IT↗

arXiv2014

Novel Realization of Adaptive Sparse Sensing with Sparse Least Mean Fourth Algorithm

Nonlinear sparse sensing (NSS) techniques have been adopted for realizing compressive sensing (CS) in many applications such as Radar imaging and sparse channel estimation. Unlike the NSS, in this paper, we propose an adaptive sparse sensing (ASS) approach using reweighted zero-attracting normalized least mean fourth (RZA-NLMF) algorithm which depends on several given parameters, i.e., reweighted factor, regularization parameter and initial step-size. First, based on the independent assumption, Cramer Rao lower bound (CRLB) is derived as for the performance comparisons. In addition, reweighted factor selection method is proposed for achieving robust estimation performance. Finally, to verify the algorithm, Monte Carlo based computer simulations are given to show that the ASS achieves much better mean square error (MSE) performance than the NSS.

cs.IT↗

arXiv2014

Adaptive MIMO Channel Estimation using Sparse Variable Step-Size NLMS Algorithms

To estimate multiple-input multiple-output (MIMO) channels, invariable step-size normalized least mean square (ISSNLMS) algorithm was applied to adaptive channel estimation (ACE). Since the MIMO channel is often described by sparse channel model due to broadband signal transmission, such sparsity can be exploited by adaptive sparse channel estimation (ASCE) methods using sparse ISS-NLMS algorithms. It is well known that step-size is a critical parameter which controls three aspects: algorithm stability, estimation performance and computational cost. The previous approaches can exploit channel sparsity but their step-sizes are keeping invariant which unable balances well the three aspects and easily cause either estimation performance loss or instability. In this paper, we propose two stable sparse variable step-size NLMS (VSS-NLMS) algorithms to improve the accuracy of MIMO channel estimators. First, ASCE for estimating MIMO channels is formulated in MIMO systems. Second, different sparse penalties are introduced to VSS-NLMS algorithm for ASCE. In addition, difference between sparse ISSNLMS algorithms and sparse VSS-NLMS ones are explained. At last, to verify the effectiveness of the proposed algorithms for ASCE, several selected simulation results are shown to prove that the proposed sparse VSS-NLMS algorithms can achieve better estimation performance than the conventional methods via mean square error (MSE) and bit error rate (BER) metrics.

cs.IT↗

arXiv2014

Affine Combination of Two Adaptive Sparse Filters for Estimating Large Scale MIMO Channels

Large scale multiple-input multiple-output (MIMO) system is considered one of promising technologies for realizing next-generation wireless communication system (5G) to increasing the degrees of freedom in space and enhancing the link reliability while considerably reducing the transmit power. However, large scale MIMO system design also poses a big challenge to traditional one-dimensional channel estimation techniques due to high complexity and curse of dimensionality problems which are caused by long delay spread as well as large number antenna. Since large scale MIMO channels often exhibit sparse or/and cluster-sparse structure, in this paper, we propose a simple affine combination of adaptive sparse channel estimation method for reducing complexity and exploiting channel sparsity in the large scale MIMO system. First, problem formulation and standard affine combination of adaptive least mean square (LMS) algorithm are introduced. Then we proposed an effective affine combination method with two sparse LMS filters and designed an approximate optimum affine combiner according to stochastic gradient search method as well. Later, to validate the proposed algorithm for estimating large scale MIMO channel, computer simulations are provided to confirm effectiveness of the proposed algorithm which can achieve better estimation performance than the conventional one as well as traditional method.

cs.IT↗

arXiv2014

Block Bayesian Sparse Learning Algorithms With Application to Estimating Channels in OFDM Systems

Cluster-sparse channels often exist in frequencyselective fading broadband communication systems. The main reason is received scattered waveform exhibits cluster structure which is caused by a few reflectors near the receiver. Conventional sparse channel estimation methods have been proposed for general sparse channel model which without considering the potential cluster-sparse structure information. In this paper, we investigate the cluster-sparse channel estimation (CS-CE) problems in the state of the art orthogonal frequencydivision multiplexing (OFDM) systems. Novel Bayesian clustersparse channel estimation (BCS-CE) methods are proposed to exploit the cluster-sparse structure by using block sparse Bayesian learning (BSBL) algorithm. The proposed methods take advantage of the cluster correlation in training matrix so that they can improve estimation performance. In addition, different from our previous method using uniform block partition information, the proposed methods can work well when the prior block partition information of channels is unknown. Computer simulations show that the proposed method has a superior performance when compared with the previous methods.

cs.IT↗

arXiv2014

RZA-NLMF algorithm based adaptive sparse sensing for realizing compressive sensing problems

Nonlinear sparse sensing (NSS) techniques have been adopted for realizing compressive sensing in many applications such as Radar imaging. Unlike the NSS, in this paper, we propose an adaptive sparse sensing (ASS) approach using reweighted zero-attracting normalized least mean fourth (RZA-NLMF) algorithm which depends on several given parameters, i.e., reweighted factor, regularization parameter and initial step-size. First, based on the independent assumption, Cramer Rao lower bound (CRLB) is derived as for the trademark of performance comparisons. In addition, reweighted factor selection method is proposed for achieving robust estimation performance. Finally, to verify the algorithm, Monte Carlo based computer simulations are given to show that the ASS achieves much better mean square error (MSE) performance than the NSS.

cs.IT↗

arXiv2013

Dense Scattering Layer Removal

We propose a new model, together with advanced optimization, to separate a thick scattering media layer from a single natural image. It is able to handle challenging underwater scenes and images taken in fog and sandstorm, both of which are with significantly reduced visibility. Our method addresses the critical issue -- this is, originally unnoticeable impurities will be greatly magnified after removing the scattering media layer -- with transmission-aware optimization. We introduce non-local structure-aware regularization to properly constrain transmission estimation without introducing the halo artifacts. A selective-neighbor criterion is presented to convert the unconventional constrained optimization problem to an unconstrained one where the latter can be efficiently solved.

cs.CV↗

arXiv2013

Global small solutions to 2-D incompressible MHD system

In this paper, we consider the global wellposedness of 2-D incompressible magneto-hydrodynamical system with small and smooth initial data. It is a coupled system between the Navier-Stokes equations and a free transport equation with an universal nonlinear coupling structure. The main difficulty of the proof lies in exploring the dissipative mechanism of the system due to the fact that there is a free transport equation in the system. To achieve this and to avoid the difficulty of propagating anisotropic regularity for the free transport equation, we first reformulate our system \eqref{1.1} in the Lagrangian coordinates \eqref{a14}. Then we employ anisotropic Littlewood-Paley analysis to establish the key {\it a priori} $L^1(\R^+; Lip(\R^2))$ estimate to the Lagrangian velocity field $Y_t$. With this estimate, we prove the global wellposedness of \eqref{a14} with smooth and small initial data by using the energy method. We emphasize that the algebraic structure of \eqref{a14} is crucial for the proofs to work. The global wellposedness of the original system \eqref{1.1} then follows by a suitable change of variables.

math.AP↗

arXiv2013

Global small solutions to three-dimensional incompressible MHD system

In this paper, we consider the global wellposedness of 3-D incompressible magneto-hydrodynamical system with small and smooth initial data. The main difficulty of the proof lies in establishing the global in time $L^1$ estimate for the velocity field due to the strong degeneracy and anisotropic spectral properties of the linearized system. To achieve this and to avoid the difficulty of propagating anisotropic regularity for the transport equation, we first write our system \eqref{B1} in the Lagrangian formulation \eqref{B11}. Then we employ anisotropic Littlewood-Paley analysis to establish the key $L^1$ in time estimates to the velocity and the gradient of the pressure in the Lagrangian coordinate. With those estimates, we prove the global wellposedness of \eqref{B11} with smooth and small initial data by using the energy method. Toward this, we will have to use the algebraic structure of \eqref{B11} in a rather crucial way. The global wellposedness of the original system \eqref{B1} then follows by a suitable change of variables together with a continuous argument. We should point out that compared with the linearized systems of 2-D MHD equations in \cite{XLZMHD1} and that of the 3-D modified MHD equations in \cite{LZ}, our linearized system \eqref{B19} here is much more degenerate, moreover, the formulation of the initial data for \eqref{B11} is more subtle than that in \cite{XLZMHD1}.

math.AP↗

arXiv2013

A Graph Theoretical Approach to Network Encoding Complexity

Consider an acyclic directed network $G$ with sources $S_1, S_2,..., S_l$ and distinct sinks $R_1, R_2,..., R_l$. For $i=1, 2,..., l$, let $c_i$ denote the min-cut between $S_i$ and $R_i$. Then, by Menger's theorem, there exists a group of $c_i$ edge-disjoint paths from $S_i$ to $R_i$, which will be referred to as a group of Menger's paths from $S_i$ to $R_i$ in this paper. Although within the same group they are edge-disjoint, the Menger's paths from different groups may have to merge with each other. It is known that by choosing Menger's paths appropriately, the number of mergings among different groups of Menger's paths is always bounded by a constant, which is independent of the size and the topology of $G$. The tightest such constant for the all the above-mentioned networks is denoted by $\mathcal{M}(c_1, c_2,..., c_2)$ when all $S_i$'s are distinct, and by $\mathcal{M}^*(c_1, c_2,..., c_2)$ when all $S_i$'s are in fact identical. It turns out that $\mathcal{M}$ and $\mathcal{M}^*$ are closely related to the network encoding complexity for a variety of networks, such as multicast networks, two-way networks and networks with multiple sessions of unicast. Using this connection, we compute in this paper some exact values and bounds in network encoding complexity using a graph theoretical approach.

cs.IT↗