IEEE Transactions on Signal Processing最新文献

{"title":"Simplicial Vector Autoregressive Models","authors":"Joshin Krishnan;Rohan Money;Baltasar Beferull-Lozano;Elvin Isufi","doi":"10.1109/TSP.2024.3503063","DOIUrl":"10.1109/TSP.2024.3503063","url":null,"abstract":"The vector autoregressive (VAR) model is extensively employed for modelling dynamic processes, yet its scalability is challenged by an overwhelming growth in parameters when dealing with several hundred time series. To overcome this issue, data relations can be leveraged as inductive priors to tackle the curse of dimensionality while still effectively modelling the time series. In this paper, we study the role of simplicial complexes as inductive biases when modelling time series defined on higher-order network structures such as edges and triangles. First, we propose two simplicial VAR models: one that models time series defined on a single simplicial level, such as edge flows, and another that jointly models multiple time series defined across different simplicial levels, ultimately capturing their spatio-temporal interdependencies. The proposed models use simplicial convolutional filters to facilitate parameter sharing and capture structure-aware spatio-temporal dependencies in a multiresolution manner. Second, we develop a joint simplicial-temporal Fourier transform to study the spectral characteristics of the models, depicting them as simplicial-temporal filters. Third, targeting streaming signals, we develop an online algorithm for learning simplicial VAR models. We prove this online learner attains a sublinear dynamic regret bound, ensuring convergence under reasonable assumptions. Finally, we corroborate the proposed approach through experiments on synthetic networks, water distribution networks, and collaborating agents. Our findings show that the proposed models attain competitive signal modelling accuracy with orders of magnitude fewer parameters than the state-of-the-art alternatives.","PeriodicalId":13330,"journal":{"name":"IEEE Transactions on Signal Processing","volume":"72 ","pages":"5454-5469"},"PeriodicalIF":4.6,"publicationDate":"2024-11-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142679112","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Directional Generation Algorithm for SAR Image Based on Azimuth-Guided Statistical Generative Adversarial Network","authors":"Guobei Peng;Ming Liu;Shichao Chen;Mingliang Tao;Yiyang Li;Mengdao Xing","doi":"10.1109/TSP.2024.3502454","DOIUrl":"10.1109/TSP.2024.3502454","url":null,"abstract":"The high cost of acquiring synthetic aperture radar (SAR) images results in the problem of insufficient data, which limits the performance of deep learning-based automatic target recognition (ATR) models. To solve this problem, a directional generation algorithm for SAR image based on azimuth-guided statistical generative adversarial network (AGSGAN) is proposed in this paper. The proposed algorithm can not only generate images with similar statistical characteristics as the real SAR images, but also control the azimuth of the generated images. Considering that the statistical characteristics of SAR images are different at different azimuth, the proposed algorithm partitions the azimuth intervals of SAR image into adaptive azimuth intervals, and the statistical characteristics of images within each adaptive azimuth interval should be similar. Then, the proposed algorithm obtains the statistical characteristics of real images by using the \u0000<inline-formula><tex-math>$G^{0}$</tex-math></inline-formula>\u0000 distribution to fit the statistical distribution of images within each adaptive azimuth interval. Finally, the random noise sampled from the fitted \u0000<inline-formula><tex-math>$G^{0}$</tex-math></inline-formula>\u0000 distribution and the serial number of adaptive azimuth interval are inputted into AGSGAN. The images that are within a specified adaptive azimuth interval and have statistical characteristics similar to the real images are generated by AGSGAN. Experimental results show that the images generated by the proposed algorithm are more realistic in statistical characteristics, and can effectively improve the recognition accuracy of the deep learning-based SAR automatic target recognition model.","PeriodicalId":13330,"journal":{"name":"IEEE Transactions on Signal Processing","volume":"72 ","pages":"5406-5421"},"PeriodicalIF":4.6,"publicationDate":"2024-11-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142673532","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Structured Directional Pruning via Perturbation Orthogonal Projection","authors":"Xiaofeng Liu;Qing Wang;Yunfeng Shao;Yanhui Geng;Yinchuan Li","doi":"10.1109/TSP.2024.3501674","DOIUrl":"10.1109/TSP.2024.3501674","url":null,"abstract":"Despite the great potential of artificial intelligence (AI), which promotes machines to mimic human intelligence in performing tasks, it requires a deep/extensive model with a sufficient number of parameters to enhance the expressive ability. This aspect often hinders the application of AI on resource-constrained devices. Structured pruning is an effective compression technique that reduces the computation of neural networks. However, it typically achieves parameter reduction at the cost of non-negligible accuracy loss, necessitating fine-tuning. This paper introduces a novel technique called Structured Directional Pruning (SDP) and its fast solver, Alternating Structured Directional Pruning (\u0000<monospace>AltSDP</monospace>\u0000). SDP is a general energy-efficient coarse-grained pruning method that enables efficient model pruning without requiring fine-tuning or expert knowledge of the desired sparsity level. Theoretical analysis confirms that the fast solver, \u0000<monospace>AltSDP</monospace>\u0000, achieves SDP asymptotically after sufficient training. Experimental results validate that \u0000<monospace>AltSDP</monospace>\u0000 reaches the same minimum valley as the vanilla optimizer, namely stochastic gradient descent (SGD), while maintaining a constant training loss. Additionally, \u0000<monospace>AltSDP</monospace>\u0000 achieves state-of-the-art pruned accuracy integrating pruning into the initial training process without the need for fine-tuning. Consequently, the newly proposed SDP, along with its fast solver \u0000<monospace>AltSDP</monospace>\u0000, can significantly facilitate the development of shrinking deep neural networks (DNNs) and enable the deployment of AI on resource-constrained devices.","PeriodicalId":13330,"journal":{"name":"IEEE Transactions on Signal Processing","volume":"72 ","pages":"5439-5453"},"PeriodicalIF":4.6,"publicationDate":"2024-11-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142673534","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Intelligent Reflecting Surface-Assisted NLOS Sensing With OFDM Signals","authors":"Jilin Wang;Jun Fang;Hongbin Li;Lei Huang","doi":"10.1109/TSP.2024.3498861","DOIUrl":"10.1109/TSP.2024.3498861","url":null,"abstract":"This work addresses the problem of intelligent reflecting surface (IRS) assisted target sensing in a non-line-of-sight (NLOS) scenario, where an IRS is employed to facilitate the radar/access point (AP) to sense the targets when the line-of-sight (LOS) path between the AP and the target is blocked by obstacles. To sense the targets, the AP transmits a train of uniformly-spaced orthogonal frequency division multiplexing (OFDM) pulses, and then perceives the targets based on the echoes from the AP-IRS-targets-IRS-AP channel. To resolve an inherent scaling ambiguity associated with IRS-assisted NLOS sensing, we propose a two-phase sensing scheme by exploiting the diversity in the illumination pattern of the IRS across two different phases. Specifically, the received echo signals from the two phases are formulated as third-order tensors. Then a canonical polyadic (CP) decomposition-based method is developed to estimate each target’s parameters including the direction of arrival (DOA), Doppler shift and time delay. Our analysis reveals that the proposed method achieves reliable NLOS sensing using a modest quantity of pulse/subcarrier resources. Simulation results are provided to show the effectiveness of the proposed method under the challenging scenario where the degrees-of-freedom provided by the AP-IRS channel are not enough for resolving the scaling ambiguity.","PeriodicalId":13330,"journal":{"name":"IEEE Transactions on Signal Processing","volume":"72 ","pages":"5322-5337"},"PeriodicalIF":4.6,"publicationDate":"2024-11-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142637343","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Joint Compression and Multiuser Equalization for Multi-Carrier Massive MIMO Systems With Decentralized Baseband Processing","authors":"Yanqing Xu;Lin Zhu;Rui Shi;Tsung-Hui Chang","doi":"10.1109/TSP.2024.3497312","DOIUrl":"10.1109/TSP.2024.3497312","url":null,"abstract":"The decentralized baseband processing (DBP) architecture is recently proposed for massive MIMO systems to reduce the interconnection cost of fronthaul links and baseband (BB) computational complexity. This paper studies the uplink multiuser equalization (MUE) problem under the DBP architecture in a multi-carrier system. Specifically, we consider a linear compression-based MUE (LC-MUE) scheme where the distributed BB units first compress the received multi-carrier signals in the frequency domain and send dimension-reduced signals to a central unit for data equalization, leading to a multi-carrier joint compression and data equalization (MC-JCDE) design problem. The MC-JCDE problem is challenging to handle because in practice the compressor is shared across multiple subcarriers, which couples the subcarrier-wise equalizers and leads to a large-dimensional problem. To develop low-complexity algorithms, we propose two new algorithms. Specifically, the first algorithm is devised based on the block coordinated descent method and non-convex alternating direction method of multipliers, which can achieve a compelling equalization accuracy and meanwhile benefit a guaranteed convergence property. The second algorithm is heuristic but enjoys further reduced complexity, which first adopts the simple carrier-wise JCDE solution, followed by a succinct aggregation step to generate a high-quality shared compressor. Simulations show that our LC-MUE scheme and proposed algorithms can approach the centralized scheme but with notably reduced fronthaul cost.","PeriodicalId":13330,"journal":{"name":"IEEE Transactions on Signal Processing","volume":"72 ","pages":"5708-5724"},"PeriodicalIF":4.6,"publicationDate":"2024-11-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142610692","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Robust Multichannel Decorrelation via Tensor Einstein Product","authors":"Shih-Yu Chang;Hsiao-Chun Wu;Guannan Liu","doi":"10.1109/TSP.2024.3495552","DOIUrl":"10.1109/TSP.2024.3495552","url":null,"abstract":"Decorrelation of multichannel signals has played a crucial preprocessing role (in prewhitening and orthogonalization) for many signal processing applications. Classical decorrelation techniques can only be applied for signal vectors. Nonetheless, many emerging big-data and sensor-network applications involve signal tensors (signal samples required to be arranged in a tensor form of arbitrary orders). Meanwhile, the existing tensor-decorrelation methods have serious limitations. First, the correlation-tensors have to be of certain particular orders. Second, the unrealistic assumption of the specific signal-tensor form, namely the canonical polyadic (CP) form, is made. Third, the correlation-tensor has to be full-rank or an extra preprocessor based on principal component analysis is required for any non-full-rank correlation tensor. To remove the aforementioned impractical limitations, we propose a novel robust approach for high-dimensional multichannel decorrelation, which can accommodate signal tensors of arbitrary orders, forms, and ranks without any need of extra preprocessor. In this work, we introduce two new tensor-decorrelation algorithms. Our first new algorithm is designed to tackle full-rank correlation-tensors and our second new algorithm is designed to tackle non-full-rank correlation-tensors. Meanwhile, we also propose a new parallel-computing paradigm to accelerate our proposed new tensor-decorrelation algorithms. To demonstrate the applicability of our proposed new scheme, we also apply our proposed new tensor-decorrelation approach to pre-whiten the tensor signals and analyze the corresponding convergence-speed and misadjustment performances of the tensor least-mean-squares (TLMS) filter. Finally, we assess the computational- and memory-complexities of our proposed new algorithms by simulations over both artificial and real data. Simulation results show that our proposed new multichannel-decorrelation algorithms outperform the existing tensor-decorrelation methods in terms of convergence speed, eigenspread, normalized mean square error (NRMSE), and estimation accuracy.","PeriodicalId":13330,"journal":{"name":"IEEE Transactions on Signal Processing","volume":"73 ","pages":"275-291"},"PeriodicalIF":4.6,"publicationDate":"2024-11-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142610690","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Gradient Networks","authors":"Shreyas Chaudhari;Srinivasa Pranav;José M.F. Moura","doi":"10.1109/TSP.2024.3496692","DOIUrl":"10.1109/TSP.2024.3496692","url":null,"abstract":"Directly parameterizing and learning gradients of functions has widespread significance, with specific applications in inverse problems, generative modeling, and optimal transport. This paper introduces gradient networks (\u0000<monospace>GradNets</monospace>\u0000): novel neural network architectures that parameterize gradients of various function classes. \u0000<monospace>GradNets</monospace>\u0000 exhibit specialized architectural constraints that ensure correspondence to gradient functions. We provide a comprehensive \u0000<monospace>GradNet</monospace>\u0000 design framework that includes methods for transforming \u0000<monospace>GradNets</monospace>\u0000 into monotone gradient networks (\u0000<monospace>mGradNets</monospace>\u0000), which are guaranteed to represent gradients of convex functions. Our results establish that our proposed \u0000<monospace>GradNet</monospace>\u0000 (and \u0000<monospace>mGradNet</monospace>\u0000) universally approximate the gradients of (convex) functions. Furthermore, these networks can be customized to correspond to specific spaces of potential functions, including transformed sums of (convex) ridge functions. Our analysis leads to two distinct \u0000<monospace>GradNet</monospace>\u0000 architectures, \u0000<monospace>GradNet-C</monospace>\u0000 and \u0000<monospace>GradNet-M</monospace>\u0000, and we describe the corresponding monotone versions, \u0000<monospace>mGradNet-C</monospace>\u0000 and \u0000<monospace>mGradNet-M</monospace>\u0000. Our empirical results demonstrate that these architectures provide efficient parameterizations and outperform existing methods by up to 15 dB in gradient field tasks and by up to 11 dB in Hamiltonian dynamics learning tasks.","PeriodicalId":13330,"journal":{"name":"IEEE Transactions on Signal Processing","volume":"73 ","pages":"324-339"},"PeriodicalIF":4.6,"publicationDate":"2024-11-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142610691","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Efficient Recovery of Sparse Graph Signals From Graph Filter Outputs","authors":"Gal Morgenstern;Tirza Routtenberg","doi":"10.1109/TSP.2024.3495225","DOIUrl":"10.1109/TSP.2024.3495225","url":null,"abstract":"This paper investigates the recovery of a node-domain sparse graph signal from the output of a graph filter. This problem, which is often referred to as the identification of the source of a diffused sparse graph signal, is seminal in the field of graph signal processing (GSP). Sparse graph signals can be used in the modeling of a variety of real-world applications in networks, such as social, biological, and power systems, and enable various GSP tasks, such as graph signal reconstruction, blind deconvolution, and sampling. In this paper, we assume double sparsity of both the graph signal and the graph topology, as well as a low-order graph filter. We propose three algorithms to reconstruct the support set of the input sparse graph signal from the graph filter output samples, leveraging these assumptions and the generalized information criterion (GIC). First, we describe the graph multiple GIC (GM-GIC) method, which is based on partitioning the dictionary elements (graph filter matrix columns) that capture information on the signal into smaller subsets. Then, the local GICs are computed for each subset and aggregated to make a global decision. Second, inspired by the well-known branch and bound (BNB) approach, we develop the graph-based branch and bound GIC (graph-BNB-GIC), and incorporate a new tractable heuristic bound tailored to the graph and graph filter characteristics. In addition, we propose the graph-based first order correction (GFOC) method, which improves existing sparse recovery methods by iteratively examining potential improvements to the GIC cost function by replacing elements from the estimated support set with elements from their one-hop neighborhood. Simulations on stochastic block model (SBM) graphs demonstrate that the proposed sparse recovery methods outperform existing techniques in terms of support set recovery and mean-squared-error (MSE), without significant computational overhead. In addition, we investigate the application of our graph-based sparse recovery methods in blind deconvolution scenarios where the graph filter is unknown. Simulations using real-world data from brain networks and pandemic diffusion analysis further demonstrate the superiority of our approach compared to graph blind deconvolution techniques.","PeriodicalId":13330,"journal":{"name":"IEEE Transactions on Signal Processing","volume":"72 ","pages":"5550-5566"},"PeriodicalIF":4.6,"publicationDate":"2024-11-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142599331","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"SAOFTRL: A Novel Adaptive Algorithmic Framework for Enhancing Online Portfolio Selection","authors":"Runhao Shi;Daniel P. Palomar","doi":"10.1109/TSP.2024.3495696","DOIUrl":"10.1109/TSP.2024.3495696","url":null,"abstract":"Strongly Adaptive meta-algorithms (SA-meta) are popular in online portfolio selection due to their resilience in adversarial environments and adaptability to market changes. However, their application is often limited by high variance in errors, stemming from calculations over small intervals with limited observations. To address this limitation, we introduce the Strongly Adaptive Optimistic Follow-the-Regularized-Leader (SAOFTRL), an advanced framework that integrates the Optimistic Follow-the-Regularized-Leader (OFTRL) strategy into SA-meta algorithms to stabilize performance. SAOFTRL is distinguished by its novel regret bound, which provides a theoretical guarantee of worst-case performance in challenging scenarios. Additionally, we reimagine SAOFTRL within a mean-variance portfolio (MVP) framework, enhanced with shrinkage estimators and adaptive rolling windows, thereby ensuring reliable average-case performance. For practical deployment, we present an efficient SAOFTRL implementation utilizing the Successive Convex Approximation (SCA) method. Empirical evaluations demonstrate SAOFTRL's superior performance and expedited convergence when compared to existing benchmarks, confirming its effectiveness and efficiency in dynamic market conditions.","PeriodicalId":13330,"journal":{"name":"IEEE Transactions on Signal Processing","volume":"72 ","pages":"5291-5305"},"PeriodicalIF":4.6,"publicationDate":"2024-11-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142599330","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Variational Inference of Structured Line Spectra Exploiting Group-Sparsity","authors":"Jakob Möderl, Erik Leitinger, Franz Pernkopf, Klaus Witrisal","doi":"10.1109/tsp.2024.3493603","DOIUrl":"https://doi.org/10.1109/tsp.2024.3493603","url":null,"abstract":"","PeriodicalId":13330,"journal":{"name":"IEEE Transactions on Signal Processing","volume":"6 1","pages":""},"PeriodicalIF":5.4,"publicationDate":"2024-11-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142597757","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}