國立虎尾科技大學 |

Statistical Advances in Simultaneous Diagonalization and Joint Object Detection.

Record Type:	Language materials, manuscript : Monograph/item
Title/Author:	Statistical Advances in Simultaneous Diagonalization and Joint Object Detection./
Author:	Xu, Yuchen.
Description:	1 online resource (156 pages)
Notes:	Source: Dissertations Abstracts International, Volume: 85-03, Section: B.
Contained By:	Dissertations Abstracts International85-03B.
Subject:	Statistics. -
Online resource:	click for full text (PQDT)
ISBN:	9798380318334

Statistical Advances in Simultaneous Diagonalization and Joint Object Detection.
Xu, Yuchen.

Statistical Advances in Simultaneous Diagonalization and Joint Object Detection. - 1 online resource (156 pages)

Source: Dissertations Abstracts International, Volume: 85-03, Section: B.

Thesis (Ph.D.)--Cornell University, 2023.

Includes bibliographical references

With the advancement of data acquisition technologies, various formats and sources of datasets become available for different analysis purposes and application fields. This work aims at pushing statistical advances for better exploiting the information from multiple data subjects with a collective point of view, especially, through simultaneous diagonalization and joint object detection. In particular, for typical machine learning input data examples including time series, images, and videos, their structurally informative nature is often accompanied with underlying configuration patterns that may be of interest to a wide range of analysis. The main contribution of this work is to cleverly utilize the jointly uncovered information to introduce novel perspectives for data explorations and interpretations.This work consists of two main types of tasks. The first one, Simultaneous Diagonalization, is designed to study common eigen-patterns in multiple possibly asymmetric matrices. Motivated by various applications, a two-sample test as well as a generalization for multiple matrices are proposed. A partial version of the test is also studied to check whether a partial set of eigenvectors is shared across samples. Additionally, a novel algorithm for the considered testing methods is introduced. Simulation studies demonstrate favorable performance for all designs. Finally, the theoretical results are utilized to decouple multiple vector auto-regression (VAR) models into univariate time series, and to test for the same stationary distribution in recurrent Markov chains with different temporal resolutions. These applications are demonstrated using macroeconomic indices of 8 countries and streamflow discharge data, respectively. The second type of task focuses on Object Detection in images and is motivated by a solid material science application on Transmission Electron Microscopy (TEM) image processing. Spatially resolved in situ TEM records the atom-scale dynamics with millisecond temporal resolution, which results in severely degraded signal-to-noise ratios (SNR). The poor SNR associated with high temporal resolution challenges the detection on atomic columns in catalyst materials undergoing structural dynamics. Two algorithms are developed sequentially to address the challenge. Blob Detection (BD) from the community of computer vision has been tailored to deal with TEM images of nanoparticle systems with high noise content present. It is demonstrated to outperform the results of other algorithms, enabling the determination of atomic column position and its intensity with a higher degree of precision, especially when SNR conditions are extremely severe. Other than classical frame-by-frame analysis, a ridge detection idea is developed as a natural extension of BD. This work harnesses temporal correlation across frames through simultaneous analysis of long image sequences, specified as a spatial-plus-temporal image tensor. Ridge detection is a classical tool to extract curvilinear features in image processing, and has great promise for trend filtering relatively stable atom-shaped objects in TEM videos. New ridge detection algorithms are defined to non-parametrically estimate explicit trajectories of atomic-level object locations as a continuous function of time. This approach is specially adapted to handle temporal analysis of objects that seemingly stochastically disappear and subsequently reappear throughout a sequence. The proposed method demonstrates high effectiveness and efficiency in simulation examples, and also delivers noticeable performance improvements in TEM implementations compared to benchmarks.

Electronic reproduction.
Ann Arbor, Mich. :
ProQuest,
2024

Mode of access: World Wide Web

ISBN: 9798380318334Subjects--Topical Terms:

556824
Statistics.
Subjects--Index Terms:

Atomic columnsIndex Terms--Genre/Form:

554714
Electronic books.

Statistical Advances in Simultaneous Diagonalization and Joint Object Detection.
LDR:05084ntm a22004097 4500 001 1151842
005 20241118135839.5
006 m o d
007 cr mn ---uuuuu
008 250605s2023 xx obm 000 0 eng d
020 $a 9798380318334
035 $a (MiAaPQ)AAI30573198
035 $a AAI30573198
040 $a MiAaPQ $b eng $c MiAaPQ $d NTU
100 1 $a Xu, Yuchen. $3 1478678
245 1 0 $a Statistical Advances in Simultaneous Diagonalization and Joint Object Detection.
264 0 $c 2023
300 $a 1 online resource (156 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertations Abstracts International, Volume: 85-03, Section: B.
500 $a Advisor: Matteson, David S.
502 $a Thesis (Ph.D.)--Cornell University, 2023.
504 $a Includes bibliographical references
520 $a With the advancement of data acquisition technologies, various formats and sources of datasets become available for different analysis purposes and application fields. This work aims at pushing statistical advances for better exploiting the information from multiple data subjects with a collective point of view, especially, through simultaneous diagonalization and joint object detection. In particular, for typical machine learning input data examples including time series, images, and videos, their structurally informative nature is often accompanied with underlying configuration patterns that may be of interest to a wide range of analysis. The main contribution of this work is to cleverly utilize the jointly uncovered information to introduce novel perspectives for data explorations and interpretations.This work consists of two main types of tasks. The first one, Simultaneous Diagonalization, is designed to study common eigen-patterns in multiple possibly asymmetric matrices. Motivated by various applications, a two-sample test as well as a generalization for multiple matrices are proposed. A partial version of the test is also studied to check whether a partial set of eigenvectors is shared across samples. Additionally, a novel algorithm for the considered testing methods is introduced. Simulation studies demonstrate favorable performance for all designs. Finally, the theoretical results are utilized to decouple multiple vector auto-regression (VAR) models into univariate time series, and to test for the same stationary distribution in recurrent Markov chains with different temporal resolutions. These applications are demonstrated using macroeconomic indices of 8 countries and streamflow discharge data, respectively. The second type of task focuses on Object Detection in images and is motivated by a solid material science application on Transmission Electron Microscopy (TEM) image processing. Spatially resolved in situ TEM records the atom-scale dynamics with millisecond temporal resolution, which results in severely degraded signal-to-noise ratios (SNR). The poor SNR associated with high temporal resolution challenges the detection on atomic columns in catalyst materials undergoing structural dynamics. Two algorithms are developed sequentially to address the challenge. Blob Detection (BD) from the community of computer vision has been tailored to deal with TEM images of nanoparticle systems with high noise content present. It is demonstrated to outperform the results of other algorithms, enabling the determination of atomic column position and its intensity with a higher degree of precision, especially when SNR conditions are extremely severe. Other than classical frame-by-frame analysis, a ridge detection idea is developed as a natural extension of BD. This work harnesses temporal correlation across frames through simultaneous analysis of long image sequences, specified as a spatial-plus-temporal image tensor. Ridge detection is a classical tool to extract curvilinear features in image processing, and has great promise for trend filtering relatively stable atom-shaped objects in TEM videos. New ridge detection algorithms are defined to non-parametrically estimate explicit trajectories of atomic-level object locations as a continuous function of time. This approach is specially adapted to handle temporal analysis of objects that seemingly stochastically disappear and subsequently reappear throughout a sequence. The proposed method demonstrates high effectiveness and efficiency in simulation examples, and also delivers noticeable performance improvements in TEM implementations compared to benchmarks.
533 $a Electronic reproduction. $b Ann Arbor, Mich. : $c ProQuest, $d 2024
538 $a Mode of access: World Wide Web
650 4 $a Statistics. $3 556824
650 4 $a Materials science. $3 557839
650 4 $a Atomic physics. $3 1182532
650 4 $a Computational physics. $3 1181955
653 $a Atomic columns
653 $a Blob Detection
653 $a Joint diagonalization
653 $a Ridge detection
653 $a Transmission Electron Microscopy
655 7 $a Electronic books. $2 local $3 554714
690 $a 0463
690 $a 0794
690 $a 0216
690 $a 0748
710 2 $a ProQuest Information and Learning Co. $3 1178819
710 2 $a Cornell University. $b Statistics. $3 1180864
773 0 $t Dissertations Abstracts International $g 85-03B.
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=30573198 $z click for full text (PQDT)