arXiv subjects

Zhiwei Xu

Publications and source records attributed to Zhiwei Xu.

At least 109 records · Page 6Linked to original sources

arXiv2021

SIDE: State Inference for Partially Observable Cooperative Multi-Agent Reinforcement Learning

As one of the solutions to the decentralized partially observable Markov decision process (Dec-POMDP) problems, the value decomposition method has achieved significant results recently. However, most value decomposition methods require the fully observable state of the environment during training, but this is not feasible in some scenarios where only incomplete and noisy observations can be obtained. Therefore, we propose a novel value decomposition framework, named State Inference for value DEcomposition (SIDE), which eliminates the need to know the global state by simultaneously seeking solutions to the two problems of optimal control and state inference. SIDE can be extended to any value decomposition method to tackle partially observable problems. By comparing with the performance of different algorithms in StarCraft II micromanagement tasks, we verified that though without accessible states, SIDE can infer the current state that contributes to the reinforcement learning process based on past local observations and even achieve superior results to many baselines in some complex scenarios.

cs.MA↗

arXiv2021

Joint CFO, Gridless Channel Estimation and Data Detection for Underwater Acoustic OFDM Systems

In this paper, we propose an iterative receiver based on gridless variational Bayesian line spectra estimation (VALSE) named JCCD-VALSE that \emph{j}ointly estimates the \emph{c}arrier frequency offset (CFO), the \emph{c}hannel with high resolution and carries out \emph{d}ata decoding. Based on a modularized point of view and motivated by the high resolution and low complexity gridless VALSE algorithm, three modules named the VALSE module, the minimum mean squared error (MMSE) module and the decoder module are built. Soft information is exchanged between the modules to progressively improve the channel estimation and data decoding accuracy. Since the delays of multipaths of the channel are treated as continuous parameters, instead of on a grid, the leakage effect is avoided. Besides, the proposed approach is a more complete Bayesian approach as all the nuisance parameters such as the noise variance, the parameters of the prior distribution of the channel, the number of paths are automatically estimated. Numerical simulations and sea test data are utilized to demonstrate that the proposed approach performs significantly better than the existing grid-based generalized approximate message passing (GAMP) based \emph{j}oint \emph{c}hannel and \emph{d}ata decoding approach (JCD-GAMP). Furthermore, it is also verified that joint processing including CFO estimation provides performance gain.

cs.IT↗

arXiv2021

MMD-MIX: Value Function Factorisation with Maximum Mean Discrepancy for Cooperative Multi-Agent Reinforcement Learning

In the real world, many tasks require multiple agents to cooperate with each other under the condition of local observations. To solve such problems, many multi-agent reinforcement learning methods based on Centralized Training with Decentralized Execution have been proposed. One representative class of work is value decomposition, which decomposes the global joint Q-value $Q_\text{jt}$ into individual Q-values $Q_a$ to guide individuals' behaviors, e.g. VDN (Value-Decomposition Networks) and QMIX. However, these baselines often ignore the randomness in the situation. We propose MMD-MIX, a method that combines distributional reinforcement learning and value decomposition to alleviate the above weaknesses. Besides, to improve data sampling efficiency, we were inspired by REM (Random Ensemble Mixture) which is a robust RL algorithm to explicitly introduce randomness into the MMD-MIX. The experiments demonstrate that MMD-MIX outperforms prior baselines in the StarCraft Multi-Agent Challenge (SMAC) environment.

cs.MA↗

arXiv2021

Learning to Coordinate via Multiple Graph Neural Networks

The collaboration between agents has gradually become an important topic in multi-agent systems. The key is how to efficiently solve the credit assignment problems. This paper introduces MGAN for collaborative multi-agent reinforcement learning, a new algorithm that combines graph convolutional networks and value-decomposition methods. MGAN learns the representation of agents from different perspectives through multiple graph networks, and realizes the proper allocation of attention between all agents. We show the amazing ability of the graph network in representation learning by visualizing the output of the graph network, and therefore improve interpretability for the actions of each agent in the multi-agent system.

cs.MA↗

arXiv2021

RANP: Resource Aware Neuron Pruning at Initialization for 3D CNNs

Although 3D Convolutional Neural Networks are essential for most learning based applications involving dense 3D data, their applicability is limited due to excessive memory and computational requirements. Compressing such networks by pruning therefore becomes highly desirable. However, pruning 3D CNNs is largely unexplored possibly because of the complex nature of typical pruning algorithms that embeds pruning into an iterative optimization paradigm. In this work, we introduce a Resource Aware Neuron Pruning (RANP) algorithm that prunes 3D CNNs at initialization to high sparsity levels. Specifically, the core idea is to obtain an importance score for each neuron based on their sensitivity to the loss function. This neuron importance is then reweighted according to the neuron resource consumption related to FLOPs or memory. We demonstrate the effectiveness of our pruning method on 3D semantic segmentation with widely used 3D-UNets on ShapeNet and BraTS'18 datasets, video classification with MobileNetV2 and I3D on UCF101 dataset, and two-view stereo matching with Pyramid Stereo Matching (PSM) network on SceneFlow dataset. In these experiments, our RANP leads to roughly 50%-95% reduction in FLOPs and 35%-80% reduction in memory with negligible loss in accuracy compared to the unpruned networks. This significantly reduces the computational resources required to train 3D CNNs. The pruned network obtained by our algorithm can also be easily scaled up and transferred to another dataset for training.

cs.CV↗

arXiv2021

Newtonalized Orthogonal Matching Pursuit for Linear Frequency Modulated Pulse Frequency Agile Radar

The linear frequency modulated (LFM) frequency agile radar (FAR) can synthesize a wide signal bandwidth through coherent processing while keeping the bandwidth of each pulse narrow. In this way, high range resolution profiles (HRRP) can be obtained without increasing the hardware system cost. Furthermore, the agility provides improved both robustness to jamming and spectrum efficiency. Motivated by the Newtonalized orthogonal matching pursuit (NOMP) for line spectral estimation problem, the NOMP for the FAR radar termed as NOMP-FAR is designed to process each coarse range bin to extract the HRRP and velocities of multiple targets, including the guide for determining the oversampling factor and the stopping criterion. In addition, it is shown that the target will cause false alarm in the nearby coarse range bins, a postprocessing algorithm is then proposed to suppress the ghost targets. Numerical simulations are conducted to demonstrate the effectiveness of NOMP-FAR.

eess.SP↗

arXiv2020

Grid-less Variational Direction of Arrival Estimation in Heteroscedastic Noise Environment

Horizontal line arrays are often employed in underwater environments to estimate the direction of arrival (DOA) of a weak signal. Conventional beamforming (CB) is robust but has wide beamwidths and high-level sidelobes. High-resolution methods such as minimum-variance distortionless response (MVDR) and subspace-based MUSIC algorithm, produce low sidelobe levels and narrow beamwidths, but are sensitive to signal mismatch and require many snapshots and the knowledge of number of sources. In addition, heteroscedastic noise where the variance varies across observations and sensors due to nonstationary environments degrades the conventional methods significantly. This paper studies DOA in heteroscedastic noise (HN) environment, where the variance of noise is varied across the snapshots and the antennas. By treating the DOAs as random variables and the nuisance parameters of the noise variance different across the snapshots and the antennas, multi-snapshot variational line spectral estimation (MVALSE) dealing with heteroscedastic noise (MVHN) is proposed, which automatically estimates the noise variance, nuisance parameters of the prior distribution, number of sources, and provides the uncertain degrees of DOA estimates. When the noise variance only varies across the snapshots or the antennas, the variants of MVHN, i.e., MVHN-S and MVHN-A, can be naturally developed. Finally, substantial numerical experiments are conducted to illustrate the proposed algorithms' performance, including a real data set in DOA application.

eess.SP↗

arXiv2020

Refining Semantic Segmentation with Superpixel by Transparent Initialization and Sparse Encoder

Although deep learning greatly improves the performance of semantic segmentation, its success mainly lies in object central areas without accurate edges. As superpixels are a popular and effective auxiliary to preserve object edges, in this paper, we jointly learn semantic segmentation with trainable superpixels. We achieve it with fully-connected layers with Transparent Initialization (TI) and efficient logit consistency using a sparse encoder. The proposed TI preserves the effects of learned parameters of pretrained networks. This avoids a significant increase of the loss of pretrained networks, which otherwise may be caused by inappropriate parameter initialization of the additional layers. Meanwhile, consistent pixel labels in each superpixel are guaranteed by logit consistency. The sparse encoder with sparse matrix operations substantially reduces both the memory requirement and the computational complexity. We demonstrated the superiority of TI over other parameter initialization methods and tested its numerical stability. The effectiveness of our proposal was validated on PASCAL VOC 2012, ADE20K, and PASCAL Context showing enhanced semantic segmentation edges. With quantitative evaluations on segmentation edges using performance ratio and F-measure, our method outperforms the state-of-the-art.

cs.CV↗

arXiv2020

Adaptive In-network Collaborative Caching for Enhanced Ensemble Deep Learning at Edge

To enhance the quality and speed of data processing and protect the privacy and security of the data, edge computing has been extensively applied to support data-intensive intelligent processing services at edge. Among these data-intensive services, ensemble learning-based services can in natural leverage the distributed computation and storage resources at edge devices to achieve efficient data collection, processing, analysis. Collaborative caching has been applied in edge computing to support services close to the data source, in order to take the limited resources at edge devices to support high-performance ensemble learning solutions. To achieve this goal, we propose an adaptive in-network collaborative caching scheme for ensemble learning at edge. First, an efficient data representation structure is proposed to record cached data among different nodes. In addition, we design a collaboration scheme to facilitate edge nodes to cache valuable data for local ensemble learning, by scheduling local caching according to a summarization of data representations from different edge nodes. Our extensive simulations demonstrate the high performance of the proposed collaborative caching scheme, which significantly reduces the learning latency and the transmission overhead.

cs.NI↗

arXiv2020

RANP: Resource Aware Neuron Pruning at Initialization for 3D CNNs

Although 3D Convolutional Neural Networks (CNNs) are essential for most learning based applications involving dense 3D data, their applicability is limited due to excessive memory and computational requirements. Compressing such networks by pruning therefore becomes highly desirable. However, pruning 3D CNNs is largely unexplored possibly because of the complex nature of typical pruning algorithms that embeds pruning into an iterative optimization paradigm. In this work, we introduce a Resource Aware Neuron Pruning (RANP) algorithm that prunes 3D CNNs at initialization to high sparsity levels. Specifically, the core idea is to obtain an importance score for each neuron based on their sensitivity to the loss function. This neuron importance is then reweighted according to the neuron resource consumption related to FLOPs or memory. We demonstrate the effectiveness of our pruning method on 3D semantic segmentation with widely used 3D-UNets on ShapeNet and BraTS'18 as well as on video classification with MobileNetV2 and I3D on UCF101 dataset. In these experiments, our RANP leads to roughly 50-95 reduction in FLOPs and 35-80 reduction in memory with negligible loss in accuracy compared to the unpruned networks. This significantly reduces the computational resources required to train 3D CNNs. The pruned network obtained by our algorithm can also be easily scaled up and transferred to another dataset for training.

cs.CV↗

arXiv2020

Fast and Differentiable Message Passing on Pairwise Markov Random Fields

Despite the availability of many Markov Random Field (MRF) optimization algorithms, their widespread usage is currently limited due to imperfect MRF modelling arising from hand-crafted model parameters and the selection of inferior inference algorithm. In addition to differentiability, the two main aspects that enable learning these model parameters are the forward and backward propagation time of the MRF optimization algorithm and its inference capabilities. In this work, we introduce two fast and differentiable message passing algorithms, namely, Iterative Semi-Global Matching Revised (ISGMR) and Parallel Tree-Reweighted Message Passing (TRWP) which are greatly sped up on a GPU by exploiting massive parallelism. Specifically, ISGMR is an iterative and revised version of the standard SGM for general pairwise MRFs with improved optimization effectiveness, and TRWP is a highly parallel version of Sequential TRW (TRWS) for faster optimization. Our experiments on the standard stereo and denoising benchmarks demonstrated that ISGMR and TRWP achieve much lower energies than SGM and Mean-Field (MF), and TRWP is two orders of magnitude faster than TRWS without losing effectiveness in optimization. We further demonstrated the effectiveness of our algorithms on end-to-end learning for semantic segmentation. Notably, our CUDA implementations are at least $7$ and $700$ times faster than PyTorch GPU implementations for forward and backward propagation respectively, enabling efficient end-to-end learning with message passing.

cs.CV↗

arXiv2020

One-bit LFMCW Radar: Spectrum Analysis and Target Detection

One-bit radar, performing signal sampling and quantization by a one-bit ADC, is a promising technology for many civilian applications due to its low-cost and low-power consumptions. In this paper, problems encountered by one-bit LFMCW radar are studied and a two-stage target detection method termed as the dimension-reduced generalized approximate message passing (DR-GAMP) approach is proposed. Firstly, the spectrum of one-bit quantized signals in a scenario with multiple targets is analyzed. It is indicated that high-order harmonics may result in false alarms (FAs) and cannot be neglected. Secondly, based on the spectrum analysis, the DR-GAMP approach is proposed to carry out target detection. Specifically, linear preprocessing methods and target predetection are firstly adopted to perform the dimension reduction, and then, the GAMP algorithm is utilized to suppress high-order harmonics and recover true targets. Finally, numerical simulations are conducted to evaluate the performance of one-bit LFMCW radar under typical parameters. It is shown that compared to the conventional radar applying linear processing methods, one-bit LFMCW radar has about $1.3$ dB performance gain when the input signal-to-noise ratios (SNRs) of targets are low. In the presence of a strong target, it has about $1.0$ dB performance loss.

eess.SP↗

arXiv2019

Multidimensional Variational Line Spectra Estimation

The fundamental multidimensional line spectral estimation problem is addressed utilizing the Bayesian methods. Motivated by the recently proposed variational line spectral estimation (VALSE) algorithm, multidimensional VALSE (MDVALSE) is developed. MDVALSE inherits the advantages of VALSE such as automatically estimating the model order, noise variance and providing uncertain degrees of frequency estimates. Compared to VALSE, the multidimensional frequencies of a single component is treated as a whole, and the probability density function is projected as independent univariate von Mises distribution to perform tractable inference. Besides, for the initialization, efficient fast Fourier transform (FFT) is adopted to significantly reduce the computation complexity of MDVALSE. Numerical results demonstrate the effectiveness of the MDVALSE, compared to state-of-art methods.

cs.IT↗

arXiv2019

Matrix Completion from Quantized Samples via Generalized Sparse Bayesian Learning

The recovery of a low rank matrix from a subset of noisy low-precision quantized samples arises in several applications such as collaborative filtering, intelligent recommendation and millimeter wave channel estimation with few bit ADCs. In this paper, a generalized sparse Bayesian learning (Gr-SBL) combining expectation propagation (EP) is proposed to solve the matrix completion (MC), which is termed as MC-Gr-SBL. The MC-Gr-SBL automatically estimates the rank, the factors and their covariance matrices, and the noise variance. In addition, MC-Gr-SBL is proposed to solve the two dimensional line spectral estimation problem by incorporating the MUSIC algorithm. Finally, substantial numerical experiments are conducted to verify the effectiveness of the proposed algorithm.

cs.IT↗

arXiv2019

Semi-Supervised Self-Growing Generative Adversarial Networks for Image Recognition

Image recognition is an important topic in computer vision and image processing, and has been mainly addressed by supervised deep learning methods, which need a large set of labeled images to achieve promising performance. However, in most cases, labeled data are expensive or even impossible to obtain, while unlabeled data are readily available from numerous free on-line resources and have been exploited to improve the performance of deep neural networks. To better exploit the power of unlabeled data for image recognition, in this paper, we propose a semi-supervised and generative approach, namely the semi-supervised self-growing generative adversarial network (SGGAN). Label inference is a key step for the success of semi-supervised learning approaches. There are two main problems in label inference: how to measure the confidence of the unlabeled data and how to generalize the classifier. We address these two problems via the generative framework and a novel convolution-block-transformation technique, respectively. To stabilize and speed up the training process of SGGAN, we employ the metric Maximum Mean Discrepancy as the feature matching objective function and achieve larger gain than the standard semi-supervised GANs (SSGANs), narrowing the gap to the supervised methods. Experiments on several benchmark datasets show the effectiveness of the proposed SGGAN on image recognition and facial attribute recognition tasks. By using the training data with only 4% labeled facial attributes, the SGGAN approach can achieve comparable accuracy with leading supervised deep learning methods with all labeled facial attributes.

cs.CV↗

arXiv2019

Multi-snapshot Newtonized Orthogonal Matching Pursuit for Line Spectrum Estimation with Multiple Measurement Vectors

In this paper, multi-snapshot Newtonized orthogonal matching pursuit (MNOMP) algorithm is proposed to deal with the line spectrum estimation with multiple measurement vectors (MMVs). MNOMP has the low computation complexity and state-of-the-art performance advantage of NOMP, and also includes two key steps: Detecting a new sinusoid on an oversampled discrete Fourier transform (DFT) grid and refining the parameters of already detected sinusoids to avoid the problem of basis mismatch. We provide a stopping criterion based on the overestimating probability of the model order. In addition, the convergence of the proposed algorithm is also proved. Finally, numerical results are conducted to show that the performance of MNOMP benefits from MMVs, and the effectiveness of MNOMP when compared against the state-of-the-art algorithms in terms of frequency estimation accuracy and computation complexity.

cs.IT↗

arXiv2018

Variational Bayesian Line Spectral Estimation with Multiple Measurement Vectors

In this paper, the line spectral estimation (LSE) problem with multiple measurement vectors (MMVs) is studied utilizing the Bayesian methods. Motivated by the recently proposed variational line spectral estimation (VALSE) method, we develop the multisnapshot VALSE (MVALSE) for multi snapshot scenarios, which is especially important in array signal processing. The MVALSE shares the advantages of the VALSE method, such as automatically estimating the model order, noise variance, weight variance, and providing the uncertain degrees of the frequency estimates. It is shown that the MVALSE can be viewed as applying the VALSE with single measurement vector (SMV) to each snapshot, and combining the intermediate data appropriately. Furthermore, the Seq-MVALSE is developed to perform sequential estimation. Finally, numerical results are conducted to demonstrate the effectiveness of the MVALSE method, compared to the state-of-the-art methods in the MMVs setting.

cs.IT↗

arXiv2018

Gridless Line Spectral Estimation with Multiple Measurement Vector via Variational Bayesian Inference

Line spectral estimation (LSE) from multi snapshot samples is studied utilizing the variational Bayesian methods. Motivated by the recently proposed variational line spectral estimation (VALSE) method for a single snapshot, we develop the multisnapshot VALSE (MVALSE) for multi snapshot scenarios, which is important for array processing. The MVALSE shares the advantages of the VALSE method, such as automatically estimating the model order, noise variance and weight variance, closed-form updates of the posterior probability density function (PDF) of the frequencies. By using multiple snapshots, MVALSE improves the recovery performance and it encodes the prior distribution naturally. Finally, numerical results demonstrate the competitive performance of the MVALSE compared to state-of-the-art methods.

eess.SP↗