IEEE Transactions on Signal Processing This is an RSS file. You can use it to subscribe to this data in your favourite RSS reader or to display this data on your own website or blog.

FedUR: Federated Learning Optimization Through Adaptive Centralized Learning Optimizers
Introducing adaptiveness to federated learning has recently ushered in a new way to optimize its convergence performance. However, adaptive learning strategies originally designed in centralized machine learning are in naїve extended to federated learning in existing works, which does not necessarily improve convergence performance and further reduce communication overhead as expected. In this paper, we fully investigate those centralized learning-based adaptive learning strategies, and propose an adaptive Federated learning algorithm targeting the model parameter Update Rule, called FedUR. Convergence upper bounds ...
Source: IEEE Transactions on Signal Processing - August 4, 2023 Category: Biomedical Engineering Source Type: research

Overcomplete Multiscale Dictionary of Slepian Functions for HEALPix on the Sphere
We present a framework for exact analytical computation of bandlimited Slepian functions for Hierarchical Equal Area iso-Latitude Pixelization (HEALPix) scheme on the sphere. Slepian functions are bandlimited eigenfunctions obtained by solving the spatial-spectral concentration problem on the sphere. Utilizing rotational symmetries between the HEALPix pixels, we employ Wigner-$D$ functions to efficiently compute the bandlimited Slepian functions at different resolutions of the HEALPix partitioning scheme. We present convergence criteria for the infinite series expansions involved in the analytical expressions, analyze the ...
Source: IEEE Transactions on Signal Processing - July 25, 2023 Category: Biomedical Engineering Source Type: research

Transferability Properties of Graph Neural Networks
Graph neural networks (GNNs) are composed of layers consisting of graph convolutions and pointwise nonlinearities. Due to their invariance and stability properties, GNNs are provably successful at learning representations from data supported on moderate-scale graphs. However, they are difficult to learn on large-scale graphs. In this article, we study the problem of training GNNs on graphs of moderate size and transferring them to large-scale graphs. We use graph limits called graphons to define limit objects for graph filters and GNNs—graphon filters and graphon neural networks (${\mathbf{W}}$NNs)—which we interpret a...
Source: IEEE Transactions on Signal Processing - July 21, 2023 Category: Biomedical Engineering Source Type: research

Uplink to Downlink Channel Covariance Transformation in FDD Systems
This paper considers frequency division duplexing massive multiple-input multiple-output systems in which the base station (BS) is equipped with either a uniform linear antenna array (ULA) or a uniform rectangular antenna array (URA). For these systems, we develop novel uplink-to-downlink channel covariance mapping schemes. These schemes can be expressed in the form of easy-to-implement affine transformations which depend only on the uplink and downlink carrier frequencies, and the BS array configurations. We derive upper bounds on the estimation errors, and we use these bounds to show that the accuracy of the proposed sch...
Source: IEEE Transactions on Signal Processing - July 21, 2023 Category: Biomedical Engineering Source Type: research

Enhanced Solutions for the Block-Term Decomposition in Rank-$(L_{r},L_{r},1)$ Terms
The block-term decompositions (BTD) represent tensors as a linear combination of low multilinear rank terms and can be explicitly related to the Canonical Polyadic decomposition (CPD). In this paper, we introduce the SECSI-BTD framework, which exploits the connection between two decompositions to estimate the block-terms of the rank-$(L_{r},L_{r},1)$ BTD. The proposed SECSI-BTD algorithm includes the initial calculation of the factor estimates using the SEmi-algebraic framework for approximate Canonical polyadic decompositions via SImultaneous Matrix Diagonalizations (SECSI), followed by clustering and refinement procedure...
Source: IEEE Transactions on Signal Processing - July 21, 2023 Category: Biomedical Engineering Source Type: research

On Causal Discovery With Convergent Cross Mapping
Convergent cross mapping is a principled causal discovery technique for signals, but its efficacy depends on a number of assumptions about the systems that generated the signals. In this work, we present a self-contained introduction to the theory of causality in state-spaces, Takens' theorem, and cross maps, and we propose conditions to check if a signal is appropriate for cross mapping. Further, we propose simple analyses based on Gaussian processes to test for these conditions in data. We show that our proposed techniques detect when convergent cross mapping may conclude erroneous results using several examples f...
Source: IEEE Transactions on Signal Processing - July 21, 2023 Category: Biomedical Engineering Source Type: research

Towards Understanding Asynchronous Advantage Actor-Critic: Convergence and Linear Speedup
Asynchronous and parallel implementation of standard reinforcement learning (RL) algorithms is a key enabler of the tremendous success of modern RL. Among many asynchronous RL algorithms, arguably the most popular and effective one is the asynchronous advantage actor-critic (A3C) algorithm. Although A3C is becoming the workhorse of RL, its theoretical properties are still not well-understood, including its non-asymptotic analysis and the performance gain of parallelism (a.k.a. linear speedup). This paper revisits the A3C algorithm and establishes its non-asymptotic convergence guarantees. Under both i.i.d. and Markovian sa...
Source: IEEE Transactions on Signal Processing - July 21, 2023 Category: Biomedical Engineering Source Type: research

Learning-Based Reconstruction of FRI Signals
Finite Rate of Innovation (FRI) sampling theory enables reconstruction of classes of continuous non-bandlimited signals that have a small number of free parameters from their low-rate discrete samples. This task is often translated into a spectral estimation problem that is solved using methods involving estimating signal subspaces, which tend to break down at a certain peak signal-to-noise ratio (PSNR). To avoid this breakdown, we consider alternative approaches that make use of information from labelled data. We propose two model-based learning methods, including deep unfolding the denoising process in spectral estimatio...
Source: IEEE Transactions on Signal Processing - July 21, 2023 Category: Biomedical Engineering Source Type: research

Robust Low-Rank Matrix Recovery via Hybrid Ordinary-Welsch Function
As a widely-used tool to resist outliers, the correntropy criterion or Welsch function has recently been exploited for robust matrix recovery. However, it down-weighs all observations including uncontaminated data. On the other hand, its implicit regularizer (IR) cannot achieve sparseness, which is a desirable property in many practical scenarios. To address these two issues, we devise a novel M-estimator called hybrid ordinary-Welsch (HOW) function, which only down-weighs the outlier-contaminated data, and the IR generated by the HOW can attain sparseness. To verify the effectiveness of the HOW function, we apply it to ro...
Source: IEEE Transactions on Signal Processing - July 21, 2023 Category: Biomedical Engineering Source Type: research

Waveform Design for Optimal PSL Under Spectral and Unimodular Constraints via Alternating Minimization
In an active sensing system, waveforms with good auto-correlations are preferred for accurate parameter estimation. Furthermore, spectral compatibility is required to avoid mutual interference between devices as the electromagnetic environment becomes increasingly crowded. Waveforms should also be unimodular due to hardware limits. In this article, a new approach to generating a unimodular sequence with an approximately optimal peak side-lobe level (PSL) in auto-correlation and adjustable stopband attenuation is proposed. The proposed method is based on alternating minimization (AM) and numerical results suggest that it ou...
Source: IEEE Transactions on Signal Processing - July 21, 2023 Category: Biomedical Engineering Source Type: research

Efficient Sampling of Non Log-Concave Posterior Distributions With Mixture of Noises
This article focuses on a challenging class of inverse problems that is often encountered in applications. The forward model is a complex non-linear black-box, potentially non-injective, whose outputs cover multiple decades in amplitude. Observations are supposed to be simultaneously damaged by additive and multiplicative noises and censorship. As needed in many applications, the aim of this work is to provide uncertainty quantification on top of parameter estimates. The resulting log-likelihood is intractable and potentially non-log-concave. An adapted Bayesian approach is proposed to provide credibility intervals along w...
Source: IEEE Transactions on Signal Processing - July 18, 2023 Category: Biomedical Engineering Source Type: research

Design of a Tap-Amplitude-Based Block Proportional Adaptive Filtering Algorithm
Proportional-type algorithms have attracted much attention because of their fast convergence ability for sparse system identification. To overcome the drawbacks of existing block proportional methods stemming from inadequate block partitioning, this article develops tap-amplitude-based block partitioning methods. In the procedure, we present two block proportional normalized least-mean-square (PNLMS) algorithms named (i) ABx-PNLMS and (ii) ABy-PNLMS. In the first algorithm, the proportional gain depends on the rank-based tap-weight block on the x-axis. In comparison, the second algorithm determines the proportional gain by...
Source: IEEE Transactions on Signal Processing - July 18, 2023 Category: Biomedical Engineering Source Type: research

Radar Target Detection via Global Optimality Conditions for Binary Quadratic Programming
This article considers the problem of radar target detection in compound Gaussian clutter background. Different from the existing detector design criteria, we propose two new detection schemes for the detection problem from the optimization perspective. Specifically, in the first scheme, the detection problem is firstly studied by introducing an auxiliary variable and transforming it into a maximum likelihood estimation problem. Under this scheme, the maximum likelihood detector and its improved version with parameter estimation are developed by using maximum likelihood criterion. In the second scheme, the detection proble...
Source: IEEE Transactions on Signal Processing - July 14, 2023 Category: Biomedical Engineering Source Type: research

Riemannian Optimization for Non-Centered Mixture of Scaled Gaussian Distributions
This article studies the statistical model of the non-centered mixture of scaled Gaussian distributions (NC-MSG). Using the Fisher-Rao information geometry associated with this distribution, we derive a Riemannian gradient descent algorithm. This algorithm is leveraged for two minimization problems. The first is the minimization of a regularized negative log-likelihood (NLL). The latter makes the trade-off between a white Gaussian distribution and the NC-MSG. Conditions on the regularization are given so that the existence of a minimum to this problem is guaranteed without assumptions on the samples. Then, the Kullback-Lei...
Source: IEEE Transactions on Signal Processing - July 14, 2023 Category: Biomedical Engineering Source Type: research

Denoising Noisy Neural Networks: A Bayesian Approach With Compensation
This article studies a fundamental problem of NoisyNNs: how to reconstruct the DNN weights from their noisy manifestations. While prior works relied exclusively on the maximum likelihood (ML) estimation, this article puts forth a denoising approach to reconstruct DNNs with the aim of maximizing the inference accuracy of the reconstructed models. The superiority of our denoiser is rigorously proven in two small-scale problems, wherein we consider a quadratic neural network function and a shallow feedforward neural network, respectively. When applied to advanced learning tasks with modern DNN architectures, our denoiser exhi...
Source: IEEE Transactions on Signal Processing - July 14, 2023 Category: Biomedical Engineering Source Type: research