In federated learning, heterogeneity in the clients' local datasets and computation speeds results in large variations in the number of local updates performed by each client in each communication round. Naive weighted aggregation of such models causes objective inconsistency, that is, the global model converges to a stationary point of a mismatched objective function which can be arbitrarily different from the true objective. This paper provides a general framework to analyze the convergence of...
#1Rui Xie(HUST: Huazhong University of Science and Technology)
#2Dengyu Hu(HUST: Huazhong University of Science and Technology)
Last. Tao Jiang(HUST: Huazhong University of Science and Technology)H-Index: 53
view all 0 authors...
To achieve high resolution, orthogonal frequency division multiplex (OFDM) radars deploy two-dimensional multiple signal classification (2D-MUSIC) in the joint range-velocity estimator. However, it is obvious that both of the signal reconstruction and 2D smoothing affect the noise statistical distribution and virtual array aperture in the joint range-velocity estimator with 2D-MUSIC. Therefore, the conventional accuracy analysis methods for the MUSIC are no longer suitable. In this paper, we pro...
Last. Zhi Ding(UC Davis: University of California, Davis)H-Index: 63
view all 2 authors...
In this work, we analyze the convergence of constant modulus algorithm (CMA) in blindly recovering multiple signals to facilitate grant-free wireless access. The CMA typically solves a non-convex problem by utilizing stochastic gradient descent. The iterative convergence of CMA can be affected by additive channel noise and finite number of samples, which is a problem not fully investigated previously. We point out the strong similarity between CMA and the Wirtinger Flow (WF) algorithm originally...
A key objective in multi-view learning is to model the information common to multiple parallel views of a class of objects/events to improve downstream tasks such as classification and clustering. In this context, two open research challenges remain; achieving scalability: how can we incorporate information from hundreds of views per event into a model and being view-agnostic: how to learn robust multi-view representations without knowledge of how these views are acquired In this work, we study ...
#2Renato J. Cintra(UFPE: Federal University of Pernambuco)H-Index: 21
Last. Sirani M. Perera(ERAU: Embry–Riddle Aeronautical University)H-Index: 4
view all 4 authors...
This paper introduces a collection of scaling methods for generating 2Npoint DCT-II approximations based on Npoint low-complexity transformations. Such scaling is based on the Hou recursive matrix factorization of the exact 2Npoint DCT-II matrix. Encompassing the widely employed Jridi-Alfalou-Meher scaling method, the proposed techniques are shown to produce DCT-II approximations that outperform the transforms resulting from the JAM scaling method according to total error energy and mea...
Distributed or decentralized estimation of covariance, and distributed principal component analysis have been introduced and studied in the signal processing community in recent years, and applications in array processing have been indicated in some detail. Inspired by these, this paper provides a detailed development of several distributed algorithms for array processing. New distributed algorithms are proposed for DOA estimation methods like root-MUSIC, total least squares-ESPRIT, and FOCUSS. ...
Sparse arrays achieve a high number of degrees-of-freedom (DOFs) by employing an equivalent virtual array signal model obtained from difference co-arrays. However, most of the existing sparse array designs fail to utilize the maximum number of DOFs due to the lag redundancies in the resulting difference co-array. In this paper, we present novel non-redundant sparse array design strategies that achieve the highest possible number of DOFs for direction-of-arrival (DOA) estimation by providing the ...
In this correspondence, we make a few corrections to the geodesic projection method of estimation fusion presented in [M. Tang, Y. Rong, J. Zhou, and X. R. Li, IEEE TSP, vol. 67, no. 2, pp. 279-292, 2019].
#2Yuejiao Sun(UCLA: University of California, Los Angeles)H-Index: 5
Last. Wotao Yin(UCLA: University of California, Los Angeles)H-Index: 70
view all 3 authors...
This paper targets developing algorithms for solving distributed machine learning problems in a communication-efcient fashion. A class of new stochastic gradient descent (SGD) approaches have been developed, which can be viewed as the stochastic generalization to the recently developed lazily aggregated gradient (LAG) method justifying the name LASG. LAG adaptively predicts the contribution of each round of communication and chooses only the signicant ones to perform. It saves communication whil...
#1Xixi Yuan(MUST: Macau University of Science and Technology)
#2Zhanchuan Cai(MUST: Macau University of Science and Technology)H-Index: 9
In this paper, we propose a new class of discontinuous orthogonal system called Walsh-U system, which is composed of piecewise polynomials of degree k in Hilbert space L^{2}[0,1]. The Walsh-U system generalizes the Walsh system from piecewise constant to piecewise polynomials, which is capable of representing discontinuous signals. Besides, the basis functions of the Walsh-U system not only possess a more concise construction, but also have stronger sparse representation capabilities than the U-...