View in EDS

Parallel and Distributed Methods for Constrained Nonconvex Optimization-Part II: Applications in Communications and Machine Learning.

Saved in:
Bibliographic Details
Title: Parallel and Distributed Methods for Constrained Nonconvex Optimization-Part II: Applications in Communications and Machine Learning.
Authors: Scutari, Gesualdo1, Facchinei, Francisco2, Lampariello, Lorenzo3, Sardellitti, Stefania4, Song, Peiran5
Source: IEEE Transactions on Signal Processing. Apr2017, Vol. 65 Issue 8, p1945-1960. 16p.
Subjects: Distributed computing, Distributed computing management, Distributed computing software, Distributed algorithms, MIMO systems
Abstract: In Part I of this paper, we proposed and analyzed a novel algorithmic framework for the minimization of a nonconvex objective function, subject to nonconvex constraints, based on inner convex approximations. This Part II is devoted to the (nontrivial) application of the framework to the following relevant large-scale problems ranging from communications to machine learning: 1) (generalizations of) the rate profile maximization in MIMO interference broadcast networks; 2) the max–min fair multicast multigroup beamforming problem in a multicell environment; and 3) a general nonconvex constrained bi-criteria formulation for $k$ -sparse variable selection in statistical learning; the two criteria are a nonconvex loss objective function, measuring the fitness of the model to data, and the latter is a nonconvex sparsity-inducing constraint in the general form of difference-of-convex (DC) functions, which allows to accomodate in a unified fashion convex and nonconvex surrogates of the $\ell _0$ function. The proposed algorithms outperform current state-of-the-art schemes for 1)–3) both theoretically and numerically. For instance, they are the first distributed schemes for the class of problems 1) and 2); and they also lead to subproblems enjoying closed form solutions. [ABSTRACT FROM PUBLISHER]
Copyright of IEEE Transactions on Signal Processing is the property of IEEE and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database: Engineering Source
Description
Abstract:In Part I of this paper, we proposed and analyzed a novel algorithmic framework for the minimization of a nonconvex objective function, subject to nonconvex constraints, based on inner convex approximations. This Part II is devoted to the (nontrivial) application of the framework to the following relevant large-scale problems ranging from communications to machine learning: 1) (generalizations of) the rate profile maximization in MIMO interference broadcast networks; 2) the max–min fair multicast multigroup beamforming problem in a multicell environment; and 3) a general nonconvex constrained bi-criteria formulation for $k$ -sparse variable selection in statistical learning; the two criteria are a nonconvex loss objective function, measuring the fitness of the model to data, and the latter is a nonconvex sparsity-inducing constraint in the general form of difference-of-convex (DC) functions, which allows to accomodate in a unified fashion convex and nonconvex surrogates of the $\ell _0$ function. The proposed algorithms outperform current state-of-the-art schemes for 1)–3) both theoretically and numerically. For instance, they are the first distributed schemes for the class of problems 1) and 2); and they also lead to subproblems enjoying closed form solutions. [ABSTRACT FROM PUBLISHER]
ISSN:1053587X
DOI:10.1109/TSP.2016.2637314