Searcharxiv⌕ Search

arXiv subjects

Xiaodong Wang

Publications and source records attributed to Xiaodong Wang.

At least 145 records · Page 8Linked to original sources
arXiv2021

Reconfigurable-intelligent-surface-assisted Downlink Transmission Design via Bayesian Optimization

This paper investigates the transmission design in the reconfigurable-intelligent-surface (RIS)-assisted downlink system. The channel state information (CSI) is usually difficult to be estimated at the base station (BS) when the RIS is not equipped with radio frequency chains. In this paper, we propose a downlink transmission framework with unknown CSI via Bayesian optimization. Since the CSI is not available at the BS, we treat the unknown objective function as the black-box function and take the beamformer, the phase shift, and the receiving filter as the input. Then the objective function is decomposed as the sum of low-dimension subfunctions to reduce the complexity. By re-expressing the power constraint of the BS in spherical coordinates, the original constraint problem is converted into an equivalent unconstrained problem. The users estimate the sum MSE of the training symbols as the objective value and feed it back to the BS. We assume a Gaussian prior of the feedback samples and the next query point is updated by minimizing the constructed acquisition function. Furthermore, this framework can also be applied to the power transfer system and fairness problems. Simulation results validate the effectiveness of the proposed transmission scheme in the downlink data transmission and power transfer.

cs.IT↗
arXiv2021

Privacy-preserving Channel Estimation in Cell-free Hybrid Massive MIMO Systems

We consider a cell-free hybrid massive multiple-input multiple-output (MIMO) system with $K$ users and $M$ access points (APs), each with $N_a$ antennas and $N_r< N_a$ radio frequency (RF) chains. When $K\ll M{N_a}$, efficient uplink channel estimation and data detection with reduced number of pilots can be performed based on low-rank matrix completion. However, such a scheme requires the central processing unit (CPU) to collect received signals from all APs, which may enable the CPU to infer the private information of user locations. We therefore develop and analyze privacy-preserving channel estimation schemes under the framework of differential privacy (DP). As the key ingredient of the channel estimator, two joint differentially private noisy matrix completion algorithms based respectively on Frank-Wolfe iteration and singular value decomposition are presented. We provide an analysis on the tradeoff between the privacy and the channel estimation error. In particular, we show that the estimation error can be mitigated while maintaining the same privacy level by increasing the payload size with fixed pilot size; and the scaling laws of both the privacy-induced and privacy-independent error components in terms of payload size are characterized. Simulation results are provided to further demonstrate the tradeoff between privacy and channel estimation performance.

cs.IT↗
arXiv2020

Unfolded Deep Neural Network (UDNN) for High Mobility Channel Estimation

High mobility channel estimation is crucial for beyond 5G (B5G) or 6G wireless communication networks. This paper is concerned with channel estimation of high mobility OFDM communication systems. First, a two-dimensional compressed sensing problem is formulated by approximately linearizing the channel as a product of an overcomplete dictionary with a sparse vector, in which the Doppler effect caused by the high mobility channel is considered. To solve the problem that the traditional compressed sensing algorithms have too many iterations and are time consuming, we propose an unfolded deep neural network (UDNN) as the fast solver, which is inspired by the structure of iterative shrinkage-thresholding algorithm (ISTA). All the parameters in UDNN (e.g. nonlinear transforms, shrinkage thresholds, measurement matrices, etc.) are learned end-to-end, rather than being hand-crafted. Experiments demonstrate that the proposed UDNN performs better than ISTA for OFDM high mobility channel estimation, while maintaining extremely fast computational speed.

cs.IT↗
arXiv2020

Understanding Training Efficiency of Deep Learning Recommendation Models at Scale

The use of GPUs has proliferated for machine learning workflows and is now considered mainstream for many deep learning models. Meanwhile, when training state-of-the-art personal recommendation models, which consume the highest number of compute cycles at our large-scale datacenters, the use of GPUs came with various challenges due to having both compute-intensive and memory-intensive components. GPU performance and efficiency of these recommendation models are largely affected by model architecture configurations such as dense and sparse features, MLP dimensions. Furthermore, these models often contain large embedding tables that do not fit into limited GPU memory. The goal of this paper is to explain the intricacies of using GPUs for training recommendation models, factors affecting hardware efficiency at scale, and learnings from a new scale-up GPU server design, Zion.

cs.AR↗
arXiv2020

ADMM-Net for Communication Interference Removal in Stepped-Frequency Radar

Complex ADMM-Net, a complex-valued neural network architecture inspired by the alternating direction method of multipliers (ADMM), is designed for interference removal in super-resolution stepped frequency radar angle-range-doppler imaging. Tailored to an uncooperative scenario wherein a MIMO radar shares spectrum with communications, the ADMM-Net recovers the radar image---which is assumed to be sparse---and simultaneously removes the communication interference, which is modeled as sparse in the frequency domain owing to spectrum underutilization. The scenario motivates an $\ell_1$-minimization problem whose ADMM iteration, in turn, undergirds the neural network design, yielding a set of generalized ADMM iterations that have learnable hyperparameters and operations. To train the network we use random data generated according to the radar and communication signal models. In numerical experiments ADMM-Net exhibits markedly lower error and computational cost than ADMM and CVX.

eess.SP↗
arXiv2020

Spectral Method for Phase Retrieval: an Expectation Propagation Perspective

Phase retrieval refers to the problem of recovering a signal $\mathbf{x}_{\star}\in\mathbb{C}^n$ from its phaseless measurements $y_i=|\mathbf{a}_i^{\mathrm{H}}\mathbf{x}_{\star}|$, where $\{\mathbf{a}_i\}_{i=1}^m$ are the measurement vectors. Many popular phase retrieval algorithms are based on the following two-step procedure: (i) initialize the algorithm based on a spectral method, (ii) refine the initial estimate by a local search algorithm (e.g., gradient descent). The quality of the spectral initialization step can have a major impact on the performance of the overall algorithm. In this paper, we focus on the model where the measurement matrix $\mathbf{A}=[\mathbf{a}_1,\ldots,\mathbf{a}_m]^{\mathrm{H}}$ has orthonormal columns, and study the spectral initialization under the asymptotic setting $m,n\to\infty$ with $m/n\toδ\in(1,\infty)$. We use the expectation propagation framework to characterize the performance of spectral initialization for Haar distributed matrices. Our numerical results confirm that the predictions of the EP method are accurate for not-only Haar distributed matrices, but also for realistic Fourier based models (e.g. the coded diffraction model). The main findings of this paper are the following: (1) There exists a threshold on $δ$ (denoted as $δ_{\mathrm{weak}}$) below which the spectral method cannot produce a meaningful estimate. We show that $δ_{\mathrm{weak}}=2$ for the column-orthonormal model. In contrast, previous results by Mondelli and Montanari show that $δ_{\mathrm{weak}}=1$ for the i.i.d. Gaussian model. (2) The optimal design for the spectral method coincides with that for the i.i.d. Gaussian model, where the latter was recently introduced by Luo, Alghamdi and Lu.

cs.IT↗
arXiv2020

Deep Learning Training in Facebook Data Centers: Design of Scale-up and Scale-out Systems

Large-scale training is important to ensure high performance and accuracy of machine-learning models. At Facebook we use many different models, including computer vision, video and language models. However, in this paper we focus on the deep learning recommendation models (DLRMs), which are responsible for more than 50% of the training demand in our data centers. Recommendation models present unique challenges in training because they exercise not only compute but also memory capacity as well as memory and network bandwidth. As model size and complexity increase, efficiently scaling training becomes a challenge. To address it we design Zion - Facebook's next-generation large-memory training platform that consists of both CPUs and accelerators. Also, we discuss the design requirements of future scale-out training systems.

cs.DC↗
arXiv2020

Multi-mode OAM Radio Waves: Generation, Angle of Arrival Estimation and Reception With UCAs

Orbital angular momentum (OAM) at radio frequency (RF) provides a novel approach of multiplexing a set of orthogonal modes on the same frequency channel to achieve high spectrum efficiencies. However, there are still big challenges in the multi-mode OAM generation, OAM antenna alignment and OAM signal reception. To solve these problems, we propose an overall scheme of the line-of-sight multi-carrier and multi-mode OAM (LoS MCMM-OAM) communication based on uniform circular arrays (UCAs). First, we verify that UCA can generate multi-mode OAM radio beam with both the RF analog synthesis method and the baseband digital synthesis method. Then, for the considered UCA-based LoS MCMM-OAM communication system, a distance and AoA estimation method is proposed based on the two-dimensional ESPRIT (2-D ESPRIT) algorithm. A salient feature of the proposed LoS MCMM-OAM and LoS MCMM-OAM-MIMO systems is that the channel matrices are completely characterized by three parameters, namely, the azimuth angle, the elevation angle and the distance, independent of the numbers of subcarriers and antennas, which significantly reduces the burden by avoiding estimating large channel matrices, as traditional MIMO-OFDM systems. After that, we propose an OAM reception scheme including the beam steering with the estimated AoA and the amplitude detection with the estimated distance. At last, the proposed methods are extended to the LoS MCMM-OAM-MIMO system equipped with uniform concentric circular arrays (UCCAs). Both mathematical analysis and simulation results validate that the proposed OAM reception scheme can eliminate the effect of the misalignment error of a practical OAM channel and approaches the performance of an ideally aligned OAM channel.

eess.SP↗
arXiv2020

Exploiting Parallelism Opportunities with Deep Learning Frameworks

State-of-the-art machine learning frameworks support a wide variety of design features to enable a flexible machine learning programming interface and to ease the programmability burden on machine learning developers. Identifying and using a performance-optimal setting in feature-rich frameworks, however, involves a non-trivial amount of performance profiling efforts and often relies on domain-specific knowledge. This paper takes a deep dive into analyzing the performance impact of key design features in a machine learning framework and quantifies the role of parallelism. The observations and insights distill into a simple set of guidelines that one can use to achieve much higher training and inference speedup. Across a diverse set of real-world deep learning models, the evaluation results show that the proposed performance tuning guidelines outperform the Intel and TensorFlow recommended settings by 1.29x and 1.34x, respectively.

cs.LG↗
arXiv2020

mmWave/THz Channel Estimation Using Frequency-Selective Atomic Norm Minimization

We propose a MIMO channel estimation method for millimeter-wave (mmWave) and terahertz (THz) systems based on frequency-selective atomic norm minimization (FS-ANM). For the strong line-of-sight property of the channel in such high-frequency bands, prior knowledge on the ranges of angles of departure/arrival (AoD/AoA) can be obtained as the prior knowledge, which can be exploited by the proposed channel estimator to improve the estimation accuracy. Simulation results show that the proposed method can achieve considerable performance gain when compared with the existing approaches without incorporating the the strong line-of-sight property.

eess.SP↗
arXiv2020

Liouville type theorems on manifolds with nonnegative curvature and strictly convex boundary

We prove some Liouville type theorems on smooth compact Riemannian manifolds with nonnegative sectional curvature and strictly convex boundary. This gives a nonlinear generalization in low dimension of the recent sharp lower bound of the first Steklov eigenvalue by Xia-Xiong and verifies partially a conjecture by the third author. As a consequence, we derive several sharp Sobolev trace inequalities on these manifolds.

math.DG↗
arXiv2020

From learning gait signatures of many individuals to reconstructing gait dynamics of one single individual

Based on the same databases, we computationally address two seemingly highly related, in fact drastically distinct, questions via computational data-driven algorithms: 1) how to precisely achieve the big task of differentiating gait signatures of many individuals? 2) how to reconstruct an individual's complex gait dynamics in full? Our brains can "effortlessly" resolve the first question, but will definitely fail in the second one. Since many fine temporal scale gait patterns surely escape our eyes. Based on accelerometers' 3D gait time series databases, we link the answers toward both questions via multiscale structural dependency within gait dynamics of our musculoskeletal system. Two types of dependency manifestations are explored. We first develop simple algorithmic computing called Principle System-State Analysis (PSSA) for the coarse dependency in implicit forms. PSSA is shown to be able to efficiently classifying among many subjects. We then develop a multiscale Local-1st-Global-2nd (L1G2) Coding Algorithm and a landmark computing algorithm. With both algorithms, we can precisely dissect rhythmic gait cycles, and then decompose each cycle into a series of cyclic gait phases. With proper color-coding and stacking, we reconstruct and represent an individual's gait dynamics via a 3D cylinder to collectively reveal universal deterministic and stochastic structural patterns on centisecond (10 milliseconds) scale across all rhythmic cycles. This 3D cylinder can serve as "passtensor" for authentication purposes related to clinical diagnoses and cybersecurity.

eess.SP↗
arXiv2020

Compressing Recurrent Neural Networks Using Hierarchical Tucker Tensor Decomposition

Recurrent Neural Networks (RNNs) have been widely used in sequence analysis and modeling. However, when processing high-dimensional data, RNNs typically require very large model sizes, thereby bringing a series of deployment challenges. Although the state-of-the-art tensor decomposition approaches can provide good model compression performance, these existing methods are still suffering some inherent limitations, such as restricted representation capability and insufficient model complexity reduction. To overcome these limitations, in this paper we propose to develop compact RNN models using Hierarchical Tucker (HT) decomposition. HT decomposition brings strong hierarchical structure to the decomposed RNN models, which is very useful and important for enhancing the representation capability. Meanwhile, HT decomposition provides higher storage and computational cost reduction than the existing tensor decomposition approaches for RNN compression. Our experimental results show that, compared with the state-of-the-art compressed RNN models, such as TT-LSTM, TR-LSTM and BT-LSTM, our proposed HT-based LSTM (HT-LSTM), consistently achieves simultaneous and significant increases in both compression ratio and test accuracy on different datasets.

cs.LG↗
arXiv2020

GEVO: GPU Code Optimization using Evolutionary Computation

GPUs are a key enabler of the revolution in machine learning and high performance computing, functioning as de facto co-processors to accelerate large-scale computation. As the programming stack and tool support have matured, GPUs have also become accessible to programmers, who may lack detailed knowledge of the underlying architecture and fail to fully leverage the GPU's computation power. GEVO (Gpu optimization using EVOlutionary computation) is a tool for automatically discovering optimization opportunities and tuning the performance of GPU kernels in the LLVM representation. GEVO uses population-based search to find edits to GPU code compiled to LLVM-IR and improves performance on desired criteria while retaining required functionality. We demonstrate that GEVO improves the execution time of the GPU programs in the Rodinia benchmark suite and the machine learning models, SVM and ResNet18, on NVIDIA Tesla P100. For the Rodinia benchmarks, GEVO improves GPU kernel runtime performance by an average of 49.48% and by as much as 412% over the fully compiler-optimized baseline. If kernel output accuracy is relaxed to tolerate up to 1% error, GEVO can find kernel variants that outperform the baseline version by an average of 51.08%. For the machine learning workloads, GEVO achieves kernel performance improvement for SVM on the MNIST handwriting recognition (3.24X) and the a9a income prediction (2.93X) datasets with no loss of model accuracy. GEVO achieves 1.79X kernel performance improvement on image classification using ResNet18/CIFAR-10, with less than 1% model accuracy reduction.

cs.NE↗
arXiv2020

Long-term scheduling and power control for wirelessly powered cell-free IoT

We investigate the long-term scheduling and power control scheme for a wirelessly powered cell-free Internet of Things (IoT) network which consists of distributed access points (APs) and large number of sensors. In each time slot, a subset of sensors are scheduled for uplink data transmission or downlink power transfer. Through asymptotic analysis, we obtain closedform expressions for the harvested energy and the achievable rates that are independent of random pilots. Then, using these expressions, we formulate a long-term scheduling and power control problem to maximize the minimum time average achievable rate among all sensors, while maintaining the battery state of each sensor higher than a predefined minimum level. Using Lyapunov optimization, the transmission mode, the active sensor set, and the power control coefficients for each time slot are jointly determined. Finally, simulation results validate the accuracy of our derived closed-form expressions and reveal that the minimum time average achievable rate is boosted significantly by the proposed scheme compare with the simple greedy transmission scheme.

eess.SP↗
arXiv2020

The Architectural Implications of Facebook's DNN-based Personalized Recommendation

The widespread application of deep learning has changed the landscape of computation in the data center. In particular, personalized recommendation for content ranking is now largely accomplished leveraging deep neural networks. However, despite the importance of these models and the amount of compute cycles they consume, relatively little research attention has been devoted to systems for recommendation. To facilitate research and to advance the understanding of these workloads, this paper presents a set of real-world, production-scale DNNs for personalized recommendation coupled with relevant performance metrics for evaluation. In addition to releasing a set of open-source workloads, we conduct in-depth analysis that underpins future system design and optimization for at-scale recommendation: Inference latency varies by 60% across three Intel server generations, batching and co-location of inferences can drastically improve latency-bounded throughput, and the diverse composition of recommendation models leads to different optimization strategies.

cs.DC↗