國立虎尾科技大學 |

Learning and Benefiting From Structured Correlations.

Record Type:	Language materials, manuscript : Monograph/item
Title/Author:	Learning and Benefiting From Structured Correlations./
Author:	Mehta, Tanay.
Description:	1 online resource (96 pages)
Notes:	Source: Dissertations Abstracts International, Volume: 85-07, Section: B.
Contained By:	Dissertations Abstracts International85-07B.
Subject:	Computer science. -
Online resource:	click for full text (PQDT)
ISBN:	9798381374995

Learning and Benefiting From Structured Correlations.
Mehta, Tanay.

Learning and Benefiting From Structured Correlations. - 1 online resource (96 pages)

Source: Dissertations Abstracts International, Volume: 85-07, Section: B.

Thesis (Ph.D.)--Northeastern University, 2024.

Includes bibliographical references

Algorithmic approaches to learning problems are essentially about discerning order within seemingly random datasets. The theoretical exploration of such algorithms seeks to formalize these concepts, employing probability distributions based on real-world data to reveal underlying structures. We consider numerous data distributions inspired by the real world, showcasing various statistical tools that enable algorithms to uncover their parameters.Two concrete datasets we consider are short-text queries and radio frequency signals. We show that with robust statistical analysis even very simple neural networks are able to learn their underlying correlations. Attention-based networks outperform LSTM architectures in short-text query tasks, and a novel generative model, AttEST, provides a theoretical foundation for the superior performance of attention-based architectures.For wireless signal classification, a combined signal processing and machine learning approach is introduced. Power spectral density and cyclostationary signal processing features, combined with neural networks, enable efficient signal classification with minimal computational resources.We then consider preferential attachment graphs from network science. A novel model, preferential interdependent networks, introduces scale-free networks that grow in conjunction with each other, parameterized by their correlation. The study of semi-correlated networks reveals their unexpected resilience, aligning with real-world network behaviors.Further exploration involves arrival order inference in preferential attachment graphs, a statistically challenging problem with implications for social networks, biological networks, and disease outbreaks. We discuss some limitations of existing probabilistic tools in solving these problems and propose future directions of research.The thesis concludes with variations of the noisy graph isomorphism problem, introducing a new algorithmic challenge-counting subgraphs in noisy preferential attachment graphs. The survey of existing results underscores the potential for algorithmic advancements in this domain.

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

Mode of access: World Wide Web

ISBN: 9798381374995Subjects--Topical Terms:

573171
Computer science.
Subjects--Index Terms:

Learning problemsIndex Terms--Genre/Form:

554714
Electronic books.

Learning and Benefiting From Structured Correlations.
LDR:03466ntm a22003857 4500 001 1143063
005 20240513061056.5
006 m o d
007 cr mn ---uuuuu
008 250605s2024 xx obm 000 0 eng d
020 $a 9798381374995
035 $a (MiAaPQ)AAI30988008
035 $a AAI30988008
040 $a MiAaPQ $b eng $c MiAaPQ $d NTU
100 1 $a Mehta, Tanay. $3 1467633
245 1 0 $a Learning and Benefiting From Structured Correlations.
264 0 $c 2024
300 $a 1 online resource (96 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: 85-07, Section: B.
500 $a Advisor: Sundaram, Ravi.
502 $a Thesis (Ph.D.)--Northeastern University, 2024.
504 $a Includes bibliographical references
520 $a Algorithmic approaches to learning problems are essentially about discerning order within seemingly random datasets. The theoretical exploration of such algorithms seeks to formalize these concepts, employing probability distributions based on real-world data to reveal underlying structures. We consider numerous data distributions inspired by the real world, showcasing various statistical tools that enable algorithms to uncover their parameters.Two concrete datasets we consider are short-text queries and radio frequency signals. We show that with robust statistical analysis even very simple neural networks are able to learn their underlying correlations. Attention-based networks outperform LSTM architectures in short-text query tasks, and a novel generative model, AttEST, provides a theoretical foundation for the superior performance of attention-based architectures.For wireless signal classification, a combined signal processing and machine learning approach is introduced. Power spectral density and cyclostationary signal processing features, combined with neural networks, enable efficient signal classification with minimal computational resources.We then consider preferential attachment graphs from network science. A novel model, preferential interdependent networks, introduces scale-free networks that grow in conjunction with each other, parameterized by their correlation. The study of semi-correlated networks reveals their unexpected resilience, aligning with real-world network behaviors.Further exploration involves arrival order inference in preferential attachment graphs, a statistically challenging problem with implications for social networks, biological networks, and disease outbreaks. We discuss some limitations of existing probabilistic tools in solving these problems and propose future directions of research.The thesis concludes with variations of the noisy graph isomorphism problem, introducing a new algorithmic challenge-counting subgraphs in noisy preferential attachment graphs. The survey of existing results underscores the potential for algorithmic advancements in this domain.
533 $a Electronic reproduction. $b Ann Arbor, Mich. : $c ProQuest, $d 2024
538 $a Mode of access: World Wide Web
650 4 $a Computer science. $3 573171
650 4 $a Electrical engineering. $3 596380
653 $a Learning problems
653 $a Radio frequency signals
653 $a Machine learning approach
653 $a Attachment graphs
655 7 $a Electronic books. $2 local $3 554714
690 $a 0984
690 $a 0544
690 $a 0800
710 2 $a ProQuest Information and Learning Co. $3 1178819
710 2 $a Northeastern University. $b Computer Science. $3 1464678
773 0 $t Dissertations Abstracts International $g 85-07B.
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=30988008 $z click for full text (PQDT)