Pub Date : 2024-08-29DOI: 10.1109/TSP.2024.3435065
Mihail Georgiev;Timothy N. Davidson
The frequency and attenuation rate of a resonance in the late-time return of a radar signal are indicative of a target's geometry and conductivity, and hence they can be used as features in a variety of filtering and classification applications. However, late-time returns are typically observed over short windows at low signal-to-noise ratios (SNRs, averaged over the window), and often in the presence of sampling jitter. This can make the estimation of these parameters difficult, even when multiple measurement shots are available. In this article, we develop a new multi-shot estimation method that is based on models for the distribution of the roots of the z-transform of the received signal. Under an additive-Gaussian-noise model, we have a closed-form expression for the root distribution in terms of the resonance parameters, and the parameters are estimated by matching the model distribution to the empirical distribution. The root distribution has a strong dependence on the frequency and attenuation rate, and leads to significantly better estimates than existing techniques at low SNRs. By developing approximate models, we extend these performance advantages to scenarios with significant sampling jitter and synchronization offsets.
雷达信号晚时回波中共振的频率和衰减率可指示目标的几何形状和传导性，因此可在各种过滤和分类应用中用作特征。然而，晚间回波通常是在信噪比（SNR）较低的短窗口内观测到的，而且往往存在采样抖动。这就给估计这些参数带来了困难，即使有多个测量镜头也是如此。在本文中，我们根据接收信号 z 变换根的分布模型，开发了一种新的多镜头估计方法。在加性高斯噪声模型下，我们用共振参数得到了根分布的闭式表达式，并通过将模型分布与经验分布相匹配来估计参数。根分布对频率和衰减率有很强的依赖性，在低信噪比情况下，其估算结果明显优于现有技术。通过开发近似模型，我们将这些性能优势扩展到具有显著采样抖动和同步偏移的场景。
{"title":"Estimating Resonances in Low-SNR Late-Time Radar Returns With Sampling Jitter","authors":"Mihail Georgiev;Timothy N. Davidson","doi":"10.1109/TSP.2024.3435065","DOIUrl":"10.1109/TSP.2024.3435065","url":null,"abstract":"The frequency and attenuation rate of a resonance in the late-time return of a radar signal are indicative of a target's geometry and conductivity, and hence they can be used as features in a variety of filtering and classification applications. However, late-time returns are typically observed over short windows at low signal-to-noise ratios (SNRs, averaged over the window), and often in the presence of sampling jitter. This can make the estimation of these parameters difficult, even when multiple measurement shots are available. In this article, we develop a new multi-shot estimation method that is based on models for the distribution of the roots of the z-transform of the received signal. Under an additive-Gaussian-noise model, we have a closed-form expression for the root distribution in terms of the resonance parameters, and the parameters are estimated by matching the model distribution to the empirical distribution. The root distribution has a strong dependence on the frequency and attenuation rate, and leads to significantly better estimates than existing techniques at low SNRs. By developing approximate models, we extend these performance advantages to scenarios with significant sampling jitter and synchronization offsets.","PeriodicalId":13330,"journal":{"name":"IEEE Transactions on Signal Processing","volume":"72 ","pages":"4651-4665"},"PeriodicalIF":4.6,"publicationDate":"2024-08-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142101380","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}
Pub Date : 2024-08-26DOI: 10.1109/TSP.2024.3450270
Konstantinos D. Polyzos;Qin Lu;Georgios B. Giannakis
Labeled data can be expensive to acquire in several application domains, including medical imaging, robotics, computer vision and wireless networks to list a few. To efficiently train machine learning models under such high labeling costs, active learning (AL) judiciously selects the most informative data instances to label on-the-fly. This active sampling process can benefit from a statistical function model, that is typically captured by a Gaussian process (GP) with well-documented merits especially in the regression task. While most GP-based AL approaches rely on a single kernel function, the present contribution advocates an ensemble of GP (EGP) models with weights adapted to the labeled data collected incrementally. Building on this novel EGP model, a suite of acquisition functions emerges based on the uncertainty and disagreement rules. An adaptively weighted ensemble of EGP-based acquisition functions is advocated to further robustify performance. Extensive tests on synthetic and real datasets in the regression task showcase the merits of the proposed EGP-based approaches with respect to the single GP-based AL alternatives.
在医疗成像、机器人、计算机视觉和无线网络等多个应用领域，获取标签数据的成本都很高。为了在如此高昂的标注成本下高效地训练机器学习模型，主动学习（AL）可以明智地选择信息量最大的数据实例进行即时标注。这种主动采样过程可以从统计函数模型中获益，该模型通常由高斯过程（GP）来捕捉，其优点有据可查，尤其是在回归任务中。大多数基于 GP 的 AL 方法都依赖于单个核函数，而本论文则主张使用权重适应增量收集的标记数据的 GP（EGP）模型集合。在这一新颖的 EGP 模型基础上，根据不确定性和分歧规则产生了一套获取函数。我们提倡基于 EGP 的自适应加权采集函数集合，以进一步提高性能。在回归任务中对合成数据集和真实数据集进行的大量测试表明，与基于 GP 的单一 AL 方法相比，所提出的基于 EGP 的方法更具优势。
{"title":"Weighted Ensembles for Adaptive Active Learning","authors":"Konstantinos D. Polyzos;Qin Lu;Georgios B. Giannakis","doi":"10.1109/TSP.2024.3450270","DOIUrl":"10.1109/TSP.2024.3450270","url":null,"abstract":"Labeled data can be expensive to acquire in several application domains, including medical imaging, robotics, computer vision and wireless networks to list a few. To efficiently train machine learning models under such high labeling costs, active learning (AL) judiciously selects the most informative data instances to label on-the-fly. This active sampling process can benefit from a statistical function model, that is typically captured by a Gaussian process (GP) with well-documented merits especially in the regression task. While most GP-based AL approaches rely on a single kernel function, the present contribution advocates an \u0000<italic>ensemble</i>\u0000 of GP (EGP) models with weights adapted to the labeled data collected incrementally. Building on this novel EGP model, a suite of acquisition functions emerges based on the uncertainty and disagreement rules. An adaptively weighted \u0000<italic>ensemble</i>\u0000 of EGP-based \u0000<italic>acquisition functions</i>\u0000 is advocated to further robustify performance. Extensive tests on synthetic and real datasets in the regression task showcase the merits of the proposed EGP-based approaches with respect to the single GP-based AL alternatives.","PeriodicalId":13330,"journal":{"name":"IEEE Transactions on Signal Processing","volume":"72 ","pages":"4178-4190"},"PeriodicalIF":4.6,"publicationDate":"2024-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142085037","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}
Pub Date : 2024-08-26DOI: 10.1109/TSP.2024.3450351
Xiaochun Niu;Ermin Wei
We study distributed optimization problems over multi-agent networks, including consensus and network flow problems. Existing distributed methods neglect the heterogeneity among agents’ computational capabilities, limiting their effectiveness. To address this, we propose DISH, a dis tributed h ybrid method that leverages system heterogeneity. DISH allows agents with higher computational capabilities or lower computational costs to perform local Newton-type updates while others adopt simpler gradient-type updates. Notably, DISH covers existing methods like EXTRA, DIGing, and ESOM-0 as special cases. To analyze DISH's performance with general update directions, we formulate distributed problems as minimax problems and introduce GRAND ( g radient- r elated a scent a n d d escent) and its alternating version, Alt-GRAND, for solving these problems. GRAND generalizes DISH to centralized minimax settings, accommodating various descent ascent update directions, including gradient-type, Newton-type, scaled gradient, and other general directions, within acute angles to the partial gradients. Theoretical analysis establishes global sublinear and linear convergence rates for GRAND and Alt-GRAND in strongly-convex-nonconcave and strongly-convex-PL settings, providing linear rates for DISH. In addition, we derive the local superlinear convergence of Newton-based variations of GRAND in centralized settings to show the potentials and limitations of Newton's method in distributed settings. Numerical experiments validate the effectiveness of our methods.
{"title":"DISH: A Distributed Hybrid Optimization Method Leveraging System Heterogeneity","authors":"Xiaochun Niu;Ermin Wei","doi":"10.1109/TSP.2024.3450351","DOIUrl":"10.1109/TSP.2024.3450351","url":null,"abstract":"We study distributed optimization problems over multi-agent networks, including consensus and network flow problems. Existing distributed methods neglect the heterogeneity among agents’ computational capabilities, limiting their effectiveness. To address this, we propose DISH, a \u0000<underline>dis</u>\u0000tributed \u0000<underline>h</u>\u0000ybrid method that leverages system heterogeneity. DISH allows agents with higher computational capabilities or lower computational costs to perform local Newton-type updates while others adopt simpler gradient-type updates. Notably, DISH covers existing methods like EXTRA, DIGing, and ESOM-0 as special cases. To analyze DISH's performance with general update directions, we formulate distributed problems as minimax problems and introduce GRAND (\u0000<underline>g</u>\u0000radient-\u0000<underline>r</u>\u0000elated \u0000<underline>a</u>\u0000scent a\u0000<underline>n</u>\u0000d \u0000<underline>d</u>\u0000escent) and its alternating version, Alt-GRAND, for solving these problems. GRAND generalizes DISH to centralized minimax settings, accommodating various descent ascent update directions, including gradient-type, Newton-type, scaled gradient, and other general directions, within acute angles to the partial gradients. Theoretical analysis establishes global sublinear and linear convergence rates for GRAND and Alt-GRAND in strongly-convex-nonconcave and strongly-convex-PL settings, providing linear rates for DISH. In addition, we derive the local superlinear convergence of Newton-based variations of GRAND in centralized settings to show the potentials and limitations of Newton's method in distributed settings. Numerical experiments validate the effectiveness of our methods.","PeriodicalId":13330,"journal":{"name":"IEEE Transactions on Signal Processing","volume":"72 ","pages":"4007-4021"},"PeriodicalIF":4.6,"publicationDate":"2024-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142085040","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}
Pub Date : 2024-08-23DOI: 10.1109/TSP.2024.3449117
Hang Zheng;Zhiguo Shi;Chengwei Zhou;Sergiy A. Vorobyov;Yujie Gu
Direction-of-arrival (DOA) estimation using deep neural networks has shown great potential for applications in complicated environments. However, conventional matrix-based deep neural networks vectorize multi-dimensional signal statistics into an excessively long input, necessitating a large number of parameters in neural layers. These parameters require substantial computational resources for training. To address the problem, we propose a resource-efficient tensorized neural network for deep tensor two-dimensional DOA estimation . In this network, the covariance tensor corresponding to the uniform rectangular array (URA) is propagated to hidden state tensors that encapsulate essential signal features. To reduce the number of trainable parameters, the feedforward propagation is formulated as inverse Tucker decomposition, compressing the parameters into inverse Tucker factors. An effective tensorized backpropagation procedure is then designed to train the compressed parameters, and the Tucker rank sequences are tuned through Bayesian optimization to ensure satisfactory network performance. Our simulation results demonstrate the superiority of the proposed tensorized deep neural network over its matrix-based counterpart. In a scenario with a $10times 10$ URA and $2$ sources, the proposed network reduces the number of trained parameters by more than $122,000$ times. Consequently, it achieves faster training speed and utilizes less GPU memory, while maintains comparable estimation accuracy and angular resolution even under non-ideal conditions and in varying scenarios.
利用深度神经网络进行到达方向（DOA）估算在复杂环境中的应用已显示出巨大潜力。然而，传统的基于矩阵的深度神经网络会将多维信号统计数据矢量化为过长的输入，这就需要在神经层中设置大量参数。这些参数的训练需要大量的计算资源。为解决这一问题，我们提出了一种用于深度张量二维 DOA 估计的资源节约型张量神经网络。在该网络中，与均匀矩形阵列（URA）相对应的协方差张量被传播到包含基本信号特征的隐藏状态张量中。为了减少可训练参数的数量，前馈传播被表述为反 Tucker 分解，将参数压缩为反 Tucker 因子。然后设计了一个有效的张量反向传播程序来训练压缩参数，并通过贝叶斯优化调整塔克秩序列，以确保令人满意的网络性能。我们的模拟结果表明，与基于矩阵的深度神经网络相比，所提出的张量深度神经网络更具优势。在一个 10 美元乘以 10 美元的 URA 和 2 美元来源的场景中，所提出的网络将训练参数的数量减少了超过 122,000 美元倍。因此，它的训练速度更快，使用的 GPU 内存更少，即使在非理想条件和不同场景下，也能保持相当的估计精度和角度分辨率。
{"title":"Deep Tensor 2-D DOA Estimation for URA","authors":"Hang Zheng;Zhiguo Shi;Chengwei Zhou;Sergiy A. Vorobyov;Yujie Gu","doi":"10.1109/TSP.2024.3449117","DOIUrl":"10.1109/TSP.2024.3449117","url":null,"abstract":"Direction-of-arrival (DOA) estimation using deep neural networks has shown great potential for applications in complicated environments. However, conventional matrix-based deep neural networks vectorize multi-dimensional signal statistics into an excessively long input, necessitating a large number of parameters in neural layers. These parameters require substantial computational resources for training. To address the problem, we propose a resource-efficient tensorized neural network for \u0000<italic>deep tensor two-dimensional DOA estimation</i>\u0000. In this network, the covariance tensor corresponding to the uniform rectangular array (URA) is propagated to hidden state tensors that encapsulate essential signal features. To reduce the number of trainable parameters, the feedforward propagation is formulated as inverse Tucker decomposition, compressing the parameters into inverse Tucker factors. An effective tensorized backpropagation procedure is then designed to train the compressed parameters, and the Tucker rank sequences are tuned through Bayesian optimization to ensure satisfactory network performance. Our simulation results demonstrate the superiority of the proposed tensorized deep neural network over its matrix-based counterpart. In a scenario with a \u0000<inline-formula><tex-math>$10times 10$</tex-math></inline-formula>\u0000 URA and \u0000<inline-formula><tex-math>$2$</tex-math></inline-formula>\u0000 sources, the proposed network reduces the number of trained parameters by more than \u0000<inline-formula><tex-math>$122,000$</tex-math></inline-formula>\u0000 times. Consequently, it achieves faster training speed and utilizes less GPU memory, while maintains comparable estimation accuracy and angular resolution even under non-ideal conditions and in varying scenarios.","PeriodicalId":13330,"journal":{"name":"IEEE Transactions on Signal Processing","volume":"72 ","pages":"4065-4080"},"PeriodicalIF":4.6,"publicationDate":"2024-08-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142045668","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}
Pub Date : 2024-08-23DOI: 10.1109/TSP.2024.3449091
Jinhao Gu;Ángel F. García-Fernández;Robert E. Firth;Lennart Svensson
This paper proposes a metric to measure the dissimilarity between graphs that may have a different number of nodes. The proposed metric extends the generalised optimal subpattern assignment (GOSPA) metric, which is a metric for sets, to graphs. The proposed graph GOSPA metric includes costs associated with node attribute errors for properly assigned nodes, missed and false nodes and edge mismatches between graphs. The computation of this metric is based on finding the optimal assignments between nodes in the two graphs, with the possibility of leaving some of the nodes unassigned. We also propose a lower bound for the metric, which is also a metric for graphs and is computable in polynomial time using linear programming. The metric is first derived for undirected unweighted graphs and it is then extended to directed and weighted graphs. The properties of the metric are demonstrated via simulated and empirical datasets.
{"title":"Graph GOSPA Metric: A Metric to Measure the Discrepancy Between Graphs of Different Sizes","authors":"Jinhao Gu;Ángel F. García-Fernández;Robert E. Firth;Lennart Svensson","doi":"10.1109/TSP.2024.3449091","DOIUrl":"10.1109/TSP.2024.3449091","url":null,"abstract":"This paper proposes a metric to measure the dissimilarity between graphs that may have a different number of nodes. The proposed metric extends the generalised optimal subpattern assignment (GOSPA) metric, which is a metric for sets, to graphs. The proposed graph GOSPA metric includes costs associated with node attribute errors for properly assigned nodes, missed and false nodes and edge mismatches between graphs. The computation of this metric is based on finding the optimal assignments between nodes in the two graphs, with the possibility of leaving some of the nodes unassigned. We also propose a lower bound for the metric, which is also a metric for graphs and is computable in polynomial time using linear programming. The metric is first derived for undirected unweighted graphs and it is then extended to directed and weighted graphs. The properties of the metric are demonstrated via simulated and empirical datasets.","PeriodicalId":13330,"journal":{"name":"IEEE Transactions on Signal Processing","volume":"72 ","pages":"4037-4049"},"PeriodicalIF":4.6,"publicationDate":"2024-08-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142045667","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}
Before the execution of many standard graph signal processing (GSP) modules, such as compression and restoration, learning of a graph that encodes pairwise (dis)similarities in data is an important precursor. In data-starved scenarios, to reduce parameterization, previous graph learning algorithms make assumptions in the nodal domain on i) graph connectivity (e.g., edge sparsity), and/or ii) edge weights (e.g., positive edges only). In this paper, given an empirical covariance matrix $bar{{mathbf{C}}}$ estimated from sparse data, we consider instead a spectral-domain assumption on the graph Laplacian matrix ${mathcal{L}}$ : the first $K$ eigenvectors (called “core” eigenvectors) ${{mathbf{u}}_{k}}$ of ${mathcal{L}}$ are pre-selected—e.g., based on domain-specific knowledge—and only the remaining eigenvectors are learned and parameterized. We first prove that, inside a Hilbert space of real symmetric matrices, the subspace ${mathcal{H}}_{mathbf{u}}^{+}$ of positive semi-definite (PSD) matrices sharing a common set of core $K$ eigenvectors ${{mathbf{u}}_{k}}$ is a convex cone. Inspired by the Gram-Schmidt procedure, we then construct an efficient operator to project a given positive definite (PD) matrix onto ${mathcal{H}}_{mathbf{u}}^{+}$ . Finally, we design a hybrid graphical lasso/projection algorithm to compute a locally optimal inverse Laplacian ${mathcal{L}}^{-1}in{mathcal{H}}_{mathbf{u}}^{+}$ given $bar{{mathbf{C}}}$ . We apply our graph learning algorithm in two practical settings: parliamentary voting interpolation and predictive transform coding in image compression. Experiments show that our algorithm outperformed existing graph learning schemes in data-starved scenarios for both synthetic data and these two settings.
{"title":"Spectral Graph Learning With Core Eigenvectors Prior via Iterative GLASSO and Projection","authors":"Saghar Bagheri;Tam Thuc Do;Gene Cheung;Antonio Ortega","doi":"10.1109/TSP.2024.3446453","DOIUrl":"10.1109/TSP.2024.3446453","url":null,"abstract":"Before the execution of many standard graph signal processing (GSP) modules, such as compression and restoration, learning of a graph that encodes pairwise (dis)similarities in data is an important precursor. In data-starved scenarios, to reduce parameterization, previous graph learning algorithms make assumptions in the nodal domain on i) graph connectivity (e.g., edge sparsity), and/or ii) edge weights (e.g., positive edges only). In this paper, given an empirical covariance matrix \u0000<inline-formula><tex-math>$bar{{mathbf{C}}}$</tex-math></inline-formula>\u0000 estimated from sparse data, we consider instead a spectral-domain assumption on the graph Laplacian matrix \u0000<inline-formula><tex-math>${mathcal{L}}$</tex-math></inline-formula>\u0000: the first \u0000<inline-formula><tex-math>$K$</tex-math></inline-formula>\u0000 eigenvectors (called “core” eigenvectors) \u0000<inline-formula><tex-math>${{mathbf{u}}_{k}}$</tex-math></inline-formula>\u0000 of \u0000<inline-formula><tex-math>${mathcal{L}}$</tex-math></inline-formula>\u0000 are pre-selected—e.g., based on domain-specific knowledge—and only the remaining eigenvectors are learned and parameterized. We first prove that, inside a Hilbert space of real symmetric matrices, the subspace \u0000<inline-formula><tex-math>${mathcal{H}}_{mathbf{u}}^{+}$</tex-math></inline-formula>\u0000 of positive semi-definite (PSD) matrices sharing a common set of core \u0000<inline-formula><tex-math>$K$</tex-math></inline-formula>\u0000 eigenvectors \u0000<inline-formula><tex-math>${{mathbf{u}}_{k}}$</tex-math></inline-formula>\u0000 is a convex cone. Inspired by the Gram-Schmidt procedure, we then construct an efficient operator to project a given positive definite (PD) matrix onto \u0000<inline-formula><tex-math>${mathcal{H}}_{mathbf{u}}^{+}$</tex-math></inline-formula>\u0000. Finally, we design a hybrid graphical lasso/projection algorithm to compute a locally optimal inverse Laplacian \u0000<inline-formula><tex-math>${mathcal{L}}^{-1}in{mathcal{H}}_{mathbf{u}}^{+}$</tex-math></inline-formula>\u0000 given \u0000<inline-formula><tex-math>$bar{{mathbf{C}}}$</tex-math></inline-formula>\u0000. We apply our graph learning algorithm in two practical settings: parliamentary voting interpolation and predictive transform coding in image compression. Experiments show that our algorithm outperformed existing graph learning schemes in data-starved scenarios for both synthetic data and these two settings.","PeriodicalId":13330,"journal":{"name":"IEEE Transactions on Signal Processing","volume":"72 ","pages":"3958-3972"},"PeriodicalIF":4.6,"publicationDate":"2024-08-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142042471","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}
Pub Date : 2024-08-22DOI: 10.1109/tsp.2024.3445606
Samy Labsir, Sara El Bouch, Alexandre Renaux, Jordi Vilà-Valls, Eric Chaumette
{"title":"Cramér-Rao Bound for Lie Group Parameter Estimation with Euclidean Observations and Unknown Covariance Matrix","authors":"Samy Labsir, Sara El Bouch, Alexandre Renaux, Jordi Vilà-Valls, Eric Chaumette","doi":"10.1109/tsp.2024.3445606","DOIUrl":"https://doi.org/10.1109/tsp.2024.3445606","url":null,"abstract":"","PeriodicalId":13330,"journal":{"name":"IEEE Transactions on Signal Processing","volume":"30 1","pages":""},"PeriodicalIF":5.4,"publicationDate":"2024-08-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142042467","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}
The Gaussian process state-space models (GPSSMs) represent a versatile class of data-driven nonlinear dynamical system models. However, the presence of numerous latent variables in GPSSM incurs unresolved issues for existing variational inference approaches, particularly under the more realistic non-mean-field (NMF) assumption, including extensive training effort, compromised inference accuracy, and infeasibility for online applications, among others. In this paper, we tackle these challenges by incorporating the ensemble Kalman filter (EnKF), a well-established model-based filtering technique, into the NMF variational inference framework to approximate the posterior distribution of the latent states. This novel marriage between EnKF and GPSSM not only eliminates the need for extensive parameterization in learning variational distributions, but also enables an interpretable, closed-form approximation of the evidence lower bound (ELBO). Moreover, owing to the streamlined parameterization via the EnKF, the new GPSSM model can be easily accommodated in online learning applications. We demonstrate that the resulting EnKF-aided online algorithm embodies a principled objective function by ensuring data-fitting accuracy while incorporating model regularizations to mitigate overfitting. We also provide detailed analysis and fresh insights for the proposed algorithms. Comprehensive evaluation across diverse real and synthetic datasets corroborates the superior learning and inference performance of our EnKF-aided variational inference algorithms compared to existing methods.
{"title":"Ensemble Kalman Filtering Meets Gaussian Process SSM for Non-Mean-Field and Online Inference","authors":"Zhidi Lin;Yiyong Sun;Feng Yin;Alexandre Hoang Thiéry","doi":"10.1109/TSP.2024.3448291","DOIUrl":"10.1109/TSP.2024.3448291","url":null,"abstract":"The Gaussian process state-space models (GPSSMs) represent a versatile class of data-driven nonlinear dynamical system models. However, the presence of numerous latent variables in GPSSM incurs unresolved issues for existing variational inference approaches, particularly under the more realistic non-mean-field (NMF) assumption, including extensive training effort, compromised inference accuracy, and infeasibility for online applications, among others. In this paper, we tackle these challenges by incorporating the ensemble Kalman filter (EnKF), a well-established model-based filtering technique, into the NMF variational inference framework to approximate the posterior distribution of the latent states. This novel marriage between EnKF and GPSSM not only eliminates the need for extensive parameterization in learning variational distributions, but also enables an interpretable, closed-form approximation of the evidence lower bound (ELBO). Moreover, owing to the streamlined parameterization via the EnKF, the new GPSSM model can be easily accommodated in online learning applications. We demonstrate that the resulting EnKF-aided online algorithm embodies a principled objective function by ensuring data-fitting accuracy while incorporating model regularizations to mitigate overfitting. We also provide detailed analysis and fresh insights for the proposed algorithms. Comprehensive evaluation across diverse real and synthetic datasets corroborates the superior learning and inference performance of our EnKF-aided variational inference algorithms compared to existing methods.","PeriodicalId":13330,"journal":{"name":"IEEE Transactions on Signal Processing","volume":"72 ","pages":"4286-4301"},"PeriodicalIF":4.6,"publicationDate":"2024-08-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142042468","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}
Pub Date : 2024-08-22DOI: 10.1109/TSP.2024.3448498
Rui Zhou;Wenqiang Pu;Ming-Yi You;Qingjiang Shi
Cooperative sensing utilizes multiple receivers dispersed across different locations, capitalizing on the advantages of multiple antennas and spatial diversity gain. This mechanism is crucial for monitoring the availability of licensed spectrum for secondary use when free from primary users. However, the efficacy of cooperative sensing relies heavily on the flawless transmission of raw data from cooperating receivers to a fusion center, a condition that may not always be fulfilled in real-world scenarios. This study investigates cooperative sensing in the context where the raw data is compromised by errors introduced during transmission, attributable to a relatively high bit error rate (BER). Consequently, the data received at the fusion center becomes incomplete and contaminated. Conventional multiantenna detectors are not adequately designed to handle such situations. To overcome this, we introduce the missing-data $t$ -distribution generalized likelihood ratio test ( $mt$ GLRT) detectors to manage such problematic data at the fusion center. The structured covariance matrices are estimated from this problematic data. Efficient optimization algorithms using the generalized expectation-maximization (GEM) method are developed accordingly. Numerical experiments corroborate the robustness of the proposed cooperative sensing methods.
{"title":"A Robust Cooperative Sensing Approach for Incomplete and Contaminated Data","authors":"Rui Zhou;Wenqiang Pu;Ming-Yi You;Qingjiang Shi","doi":"10.1109/TSP.2024.3448498","DOIUrl":"10.1109/TSP.2024.3448498","url":null,"abstract":"Cooperative sensing utilizes multiple receivers dispersed across different locations, capitalizing on the advantages of multiple antennas and spatial diversity gain. This mechanism is crucial for monitoring the availability of licensed spectrum for secondary use when free from primary users. However, the efficacy of cooperative sensing relies heavily on the flawless transmission of raw data from cooperating receivers to a fusion center, a condition that may not always be fulfilled in real-world scenarios. This study investigates cooperative sensing in the context where the raw data is compromised by errors introduced during transmission, attributable to a relatively high bit error rate (BER). Consequently, the data received at the fusion center becomes incomplete and contaminated. Conventional multiantenna detectors are not adequately designed to handle such situations. To overcome this, we introduce the missing-data \u0000<inline-formula><tex-math>$t$</tex-math></inline-formula>\u0000-distribution generalized likelihood ratio test (\u0000<inline-formula><tex-math>$mt$</tex-math></inline-formula>\u0000GLRT) detectors to manage such problematic data at the fusion center. The structured covariance matrices are estimated from this problematic data. Efficient optimization algorithms using the generalized expectation-maximization (GEM) method are developed accordingly. Numerical experiments corroborate the robustness of the proposed cooperative sensing methods.","PeriodicalId":13330,"journal":{"name":"IEEE Transactions on Signal Processing","volume":"72 ","pages":"3945-3957"},"PeriodicalIF":4.6,"publicationDate":"2024-08-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142042470","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}
Pub Date : 2024-08-20DOI: 10.1109/TSP.2024.3446512
João B. O. Souza Filho;Paulo S. R. Diniz
Kernel principal component analysis (KPCA) is a powerful tool for nonlinear feature extraction, but its standard formulation is not well-suited for streaming data. Although there are efficient online KPCA solutions, there is a gap in the literature regarding genuinely sparse online KPCA algorithms. This paper introduces a novel, fast, and accurate online fixed-point algorithm designed for sparse kernel principal component extraction. Utilizing a two-level sparsifying strategy, the proposed algorithm efficiently handles streaming data and large datasets within minimal computing and memory requirements, achieving both higher accuracy and sparser components compared to existing online KPCA methods.
{"title":"A Sparse Fixed-Point Online KPCA Extraction Algorithm","authors":"João B. O. Souza Filho;Paulo S. R. Diniz","doi":"10.1109/TSP.2024.3446512","DOIUrl":"10.1109/TSP.2024.3446512","url":null,"abstract":"Kernel principal component analysis (KPCA) is a powerful tool for nonlinear feature extraction, but its standard formulation is not well-suited for streaming data. Although there are efficient online KPCA solutions, there is a gap in the literature regarding genuinely sparse online KPCA algorithms. This paper introduces a novel, fast, and accurate online fixed-point algorithm designed for sparse kernel principal component extraction. Utilizing a two-level sparsifying strategy, the proposed algorithm efficiently handles streaming data and large datasets within minimal computing and memory requirements, achieving both higher accuracy and sparser components compared to existing online KPCA methods.","PeriodicalId":13330,"journal":{"name":"IEEE Transactions on Signal Processing","volume":"72 ","pages":"4604-4617"},"PeriodicalIF":4.6,"publicationDate":"2024-08-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142021983","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}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：