東華大學圖書館 |

語系: 繁體中文

說明(常見問題)

回圖書館首頁

手機版館藏查詢

登入

回首頁

切換: 標籤 | MARC模式 | ISBD

Detection Limits and Fluctuation Res...

El Abidi, Ahmed El Alaoui.

FindBook

Google Book

Amazon

博客來

Detection Limits and Fluctuation Results in Some Spiked Random Matrix Models and Pooling of Discrete Data.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Detection Limits and Fluctuation Results in Some Spiked Random Matrix Models and Pooling of Discrete Data./
作者:	El Abidi, Ahmed El Alaoui.
出版者:	Ann Arbor : ProQuest Dissertations & Theses, : 2018,
面頁冊數:	191 p.
附註:	Source: Dissertations Abstracts International, Volume: 80-03, Section: B.
Contained By:	Dissertations Abstracts International80-03B.
標題:	Statistical physics. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=10816525
ISBN:	9780438324732

Detection Limits and Fluctuation Results in Some Spiked Random Matrix Models and Pooling of Discrete Data.
El Abidi, Ahmed El Alaoui.

Detection Limits and Fluctuation Results in Some Spiked Random Matrix Models and Pooling of Discrete Data. - Ann Arbor : ProQuest Dissertations & Theses, 2018 - 191 p.

Source: Dissertations Abstracts International, Volume: 80-03, Section: B.

Thesis (Ph.D.)--University of California, Berkeley, 2018.

This item must not be sold to any third party vendors.

In this thesis we examine the fundamental limits of detecting and recovering a weak signal hidden in a large amount of noise. We will consider two problems. The first one pertains to principal component analysis, and is about estimating and detecting the presence of a structured low-rank signal buried inside a large noise matrix. Prominent models of this problem include the so-called spiked or deformed ensembles from random matrix theory, which are distributions over matrices of the form "signal + noise." It is known in these models that the top eigenpair of the data matrix becomes indicative of the presence of this signal, or "spike", when and only when its strength is above a certain "spectral" threshold. A natural question is then whether it is possible to identify the spike, or even tell if it's really present in the data below this threshold. In the first part of this thesis, we will completely characterize the fundamental limits of the detection and estimation of the spike. The analysis leading to this characterization crucially relies on a recently discovered connection of this statistical problem with the mean-field theory of spin glasses. This connection provides the necessary tools to obtain precise control on the behavior of the posterior distribution of the spike as well as the fluctuations of the associated likelihood ratio process. The second problem we consider is that of pooling, and is about recovering a discrete signal, i.e., whose entries only take can a finite number of values, given several observation that are constrained in a combinatorial way. More precisely, in a manner akin to compressed sensing, observations of this signal can be obtained in the form of histograms: pooled measurements counting the occurrence of each symbol across subsets of the variables. We ask what is the minimal number of random measurements of this sort it takes to reconstruct the signal. In the second part of this thesis, we determine sharp upper and lower bounds on the minimal number of measurements for the signal to be essentially unique, thus in principle recoverable from the observed data. We then provide an efficient algorithm to extract it, and show that this strategy is successful only when the number of measurements is much larger than this uniqueness threshold.

ISBN: 9780438324732Subjects--Topical Terms:

536281
Statistical physics.

Detection Limits and Fluctuation Results in Some Spiked Random Matrix Models and Pooling of Discrete Data.
LDR:03445nmm a2200337 4500 001 2210401
005 20191121124215.5
008 201008s2018 ||||||||||||||||| ||eng d
020 $a 9780438324732
035 $a (MiAaPQ)AAI10816525
035 $a (MiAaPQ)berkeley:17857
035 $a AAI10816525
040 $a MiAaPQ $c MiAaPQ
100 1 $a El Abidi, Ahmed El Alaoui. $3 3437544
245 1 0 $a Detection Limits and Fluctuation Results in Some Spiked Random Matrix Models and Pooling of Discrete Data.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2018
300 $a 191 p.
500 $a Source: Dissertations Abstracts International, Volume: 80-03, Section: B.
500 $a Publisher info.: Dissertation/Thesis.
500 $a Advisor: Jordan, Michael I.
502 $a Thesis (Ph.D.)--University of California, Berkeley, 2018.
506 $a This item must not be sold to any third party vendors.
520 $a In this thesis we examine the fundamental limits of detecting and recovering a weak signal hidden in a large amount of noise. We will consider two problems. The first one pertains to principal component analysis, and is about estimating and detecting the presence of a structured low-rank signal buried inside a large noise matrix. Prominent models of this problem include the so-called spiked or deformed ensembles from random matrix theory, which are distributions over matrices of the form "signal + noise." It is known in these models that the top eigenpair of the data matrix becomes indicative of the presence of this signal, or "spike", when and only when its strength is above a certain "spectral" threshold. A natural question is then whether it is possible to identify the spike, or even tell if it's really present in the data below this threshold. In the first part of this thesis, we will completely characterize the fundamental limits of the detection and estimation of the spike. The analysis leading to this characterization crucially relies on a recently discovered connection of this statistical problem with the mean-field theory of spin glasses. This connection provides the necessary tools to obtain precise control on the behavior of the posterior distribution of the spike as well as the fluctuations of the associated likelihood ratio process. The second problem we consider is that of pooling, and is about recovering a discrete signal, i.e., whose entries only take can a finite number of values, given several observation that are constrained in a combinatorial way. More precisely, in a manner akin to compressed sensing, observations of this signal can be obtained in the form of histograms: pooled measurements counting the occurrence of each symbol across subsets of the variables. We ask what is the minimal number of random measurements of this sort it takes to reconstruct the signal. In the second part of this thesis, we determine sharp upper and lower bounds on the minimal number of measurements for the signal to be essentially unique, thus in principle recoverable from the observed data. We then provide an efficient algorithm to extract it, and show that this strategy is successful only when the number of measurements is much larger than this uniqueness threshold.
590 $a School code: 0028.
650 4 $a Statistical physics. $3 536281
650 4 $a Mathematics. $3 515831
650 4 $a Statistics. $3 517247
690 $a 0217
690 $a 0405
690 $a 0463
710 2 $a University of California, Berkeley. $b Electrical Engineering & Computer Sciences. $3 1671057
773 0 $t Dissertations Abstracts International $g 80-03B.
790 $a 0028
791 $a Ph.D.
792 $a 2018
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=10816525