IEEE Transactions on Signal Processing - new TOC TOC Alert for Publication# 78
- A New Inexact Proximal Linear Algorithm With Adaptive Stopping Criteria for Robust Phase Retrievalel febrero 19, 2024 a las 1:18 pm
This paper considers the robust phase retrieval problem, which can be cast as a nonsmooth and nonconvex optimization problem. We propose a new inexact proximal linear algorithm with the subproblem being solved inexactly. Our contributions are two adaptive stopping criteria for the subproblem. The convergence behavior of the proposed methods is analyzed. Through experiments on both synthetic and real datasets, we demonstrate that our methods are much more efficient than existing methods, such as the original proximal linear algorithm and the subgradient method.
- Precoder Design for Massive MIMO Downlink With Matrix Manifold Optimizationel febrero 12, 2024 a las 1:20 pm
We investigate the weighted sum-rate (WSR) maximization linear precoder design for massive multiple-input multiple-output (MIMO) downlink. We consider a single-cell system with multiple users and propose a unified matrix manifold optimization framework applicable to total power constraint (TPC), per-user power constraint (PUPC) and per-antenna power constraint (PAPC). We prove that the precoders under TPC, PUPC and PAPC are on distinct Riemannian submanifolds, and transform the constrained problems in Euclidean space to unconstrained ones on manifolds. In accordance with this, we derive Riemannian ingredients, including orthogonal projection, Riemannian gradient, Riemannian Hessian, retraction and vector transport, which are needed for precoder design in the matrix manifold framework. Then, Riemannian design methods using Riemannian steepest descent, Riemannian conjugate gradient and Riemannian trust region are provided to design the WSR-maximization precoders under TPC, PUPC or PAPC. Riemannian methods do not involve the inverses of the large dimensional matrices during the iterations, reducing the computational complexities of the algorithms. Complexity analyses and performance simulations demonstrate the advantages of the proposed precoder design.
- Distributed Continual Learning With CoCoA in High-Dimensional Linear Regressionel febrero 5, 2024 a las 1:16 pm
We consider estimation under scenarios where the signals of interest exhibit change of characteristics over time. In particular, we consider the continual learning problem where different tasks, e.g., data with different distributions, arrive sequentially and the aim is to perform well on the newly arrived task without performance degradation on the previously seen tasks. In contrast to the continual learning literature focusing on the centralized setting, we investigate the problem from a distributed estimation perspective. We consider the well-established distributed learning algorithm CoCoA, which distributes the model parameters and the corresponding features over the network. We provide exact analytical characterization for the generalization error of CoCoA under continual learning for linear regression in a range of scenarios, where overparameterization is of particular interest. These analytical results characterize how the generalization error depends on the network structure, the task similarity and the number of tasks, and show how these dependencies are intertwined. In particular, our results show that the generalization error can be significantly reduced by adjusting the network size, where the most favorable network size depends on task similarity and the number of tasks. We present numerical results verifying the theoretical analysis and illustrate the continual learning performance of CoCoA with a digit classification task.
- Probe: Learning Users’ Personal Projection Bias in Inter-Temporal Choicesel febrero 5, 2024 a las 1:16 pm
Inter-temporal choices involve making decisions that require weighing costs in the present against benefits in the future. One specific type of inter-temporal choice is the decision between purchasing an individual item at full price or opting for a bundle including that item at a discounted price. Previous works assume that users have accurate expectations of factors involved in these decisions. However, in reality, users’ perceptions of these factors are often biased, leading to irrational and suboptimal decision-making. In this work, we focus on two commonly observed biases: the projection bias and the reference-point effect. To address these biases, we propose a novel bias-embedded preference model called Probe. Probe introduces prospect theory from behavioral economics to model these biases, which incorporates a weight function to capture the projection bias and a value function to account for the reference-point effect. We propose a new method to learn users’ individual projection biases from historical records and present four methods to estimate the reference point within the value function. We theoretically analyze the impact of projection bias on bundle pricing strategies. Through experimental results, we show that the proposed Probe model outperforms existing methods and leads to a better understanding of users’ behaviors in bundle purchases. This investigation can facilitate a deeper comprehension of users’ decision-making processes, enable personalized services, and help sellers design better product bundling strategies.
- A New Approach for Graph Signal Separation Based on Smoothnessel febrero 5, 2024 a las 1:16 pm
Blind source separation (BSS) is a signal processing subject that has recently been extended to graph signals. Graph signals that are smooth on their own graphs provide an opportunity to separate them from their summation by knowing their underlying graphs, which is different from the conventional BSS that requires at least two mixtures of source signals. In this paper, we introduce an approach to separate smooth graph signals whose energy is concentrated on their first frequency components. This approach tries to decompose the summation signal into signals that are as smooth as possible on their underlying graphs and non-smooth on the other graphs. Moreover, in the case that the number of source signals is two, the uniqueness of our separation approach is shown, up to the uncertainty of the average value of the signals. Furthermore, we interpret the solution of our approach in the case of complement graphs by deriving exact error formulas. Finally, simulations demonstrate the efficiency of the proposed approach and its superiority over other approaches in this setting.