國立高雄大學圖資館 |

Language: English

Back

Inference and Optimization over Netw...

Carnegie Mellon University.

Inference and Optimization over Networks: Communication Efficiency and Optimality.

Record Type:	Electronic resources : Monograph/item
Title/Author:	Inference and Optimization over Networks: Communication Efficiency and Optimality.
Author:	Sahu, Anit Kumar.
Published:	Ann Arbor : ProQuest Dissertations & Theses, 2018
Description:	317 p.
Notes:	Source: Dissertation Abstracts International, Volume: 80-04(E), Section: B.
Notes:	Adviser: Soummya Kar.
Contained By:	Dissertation Abstracts International80-04B(E).
Subject:	Electrical engineering.
Online resource:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=10980276
ISBN:	9780438719989

Inference and Optimization over Networks: Communication Efficiency and Optimality.
Sahu, Anit Kumar.

Inference and Optimization over Networks: Communication Efficiency and Optimality. - Ann Arbor : ProQuest Dissertations & Theses, 2018 - 317 p.

Source: Dissertation Abstracts International, Volume: 80-04(E), Section: B.

Thesis (Ph.D.)--Carnegie Mellon University, 2018.

We study distributed inference, learning and optimization in scenarios which involve networked entities in time-varying and random networks, which are ad-hoc in nature. In this thesis, we propose distributed recursive algorithms where the networked entities simultaneously incorporate locally sensed information and information obtained from the neighborhood. The class of distributed algorithms proposed in this thesis encompasses distributed estimation, distributed composite hypothesis testing and distributed optimization. The central theme of the scenarios considered involve systems constrained by limited on board batteries and hence constrained by limited sensing, computation and extremely limited communication resources. A typical example of such resource constrained scenarios being distributed data-parallel machine learning systems, in which the individual entities are commodity devices such as cellphones.

ISBN: 9780438719989Subjects--Topical Terms:

454503
Electrical engineering.

Inference and Optimization over Networks: Communication Efficiency and Optimality.
LDR:03260nmm a2200349 4500 001 547658
005 20190513114600.5
008 190715s2018 ||||||||||||||||| ||eng d
020 $a 9780438719989
035 $a (MiAaPQ)AAI10980276
035 $a (MiAaPQ)cmu:10326
035 $a AAI10980276
040 $a MiAaPQ $c MiAaPQ
100 1 $a Sahu, Anit Kumar. $3 827037
245 1 0 $a Inference and Optimization over Networks: Communication Efficiency and Optimality.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2018
300 $a 317 p.
500 $a Source: Dissertation Abstracts International, Volume: 80-04(E), Section: B.
500 $a Adviser: Soummya Kar.
502 $a Thesis (Ph.D.)--Carnegie Mellon University, 2018.
520 $a We study distributed inference, learning and optimization in scenarios which involve networked entities in time-varying and random networks, which are ad-hoc in nature. In this thesis, we propose distributed recursive algorithms where the networked entities simultaneously incorporate locally sensed information and information obtained from the neighborhood. The class of distributed algorithms proposed in this thesis encompasses distributed estimation, distributed composite hypothesis testing and distributed optimization. The central theme of the scenarios considered involve systems constrained by limited on board batteries and hence constrained by limited sensing, computation and extremely limited communication resources. A typical example of such resource constrained scenarios being distributed data-parallel machine learning systems, in which the individual entities are commodity devices such as cellphones.
520 $a Due to the inherent ad-hoc nature of the aforementioned setups, in conjunction with random environments render these setups central coordinator-less. Keeping in mind the resource constrained nature of such setups, we propose distributed inference and optimization algorithms which characterize the interplay between communication, computation and optimality, while allowing for heterogeneity among clients in terms of objectives, data collection and statistical dependencies.
520 $a With massive data, models for learning and optimization have been getting more and more complex to the extent of being almost analytically intractable. In such models, obtaining gradients for the associated loss function is very expensive and potentially intractable due to the lack of a closed form for the loss function. A major thrust of this thesis is gradient free zeroth order optimization which encompasses distributed setups which exhibit data parallelism and also potentially analytically intractable loss functions. On top of gradient free optimization, in this thesis we also study projection free zeroth order methods for constrained optimization.
520 $a The techniques developed in this thesis are generic and are of independent interest in classical fields such as stochastic approximation, statistical decision theory and optimization.
590 $a School code: 0041.
650 4 $a Electrical engineering. $3 454503
650 4 $a Mathematics. $3 184409
650 4 $a Statistics. $3 182057
690 $a 0544
690 $a 0405
690 $a 0463
710 2 $a Carnegie Mellon University. $b Electrical and Computer Engineering. $3 660346
773 0 $t Dissertation Abstracts International $g 80-04B(E).
790 $a 0041
791 $a Ph.D.
792 $a 2018
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=10980276