國立虎尾科技大學 |

Algorithmic Fairness in Graph Machine Learning : = Explanation, Optimization, and Certification.

Record Type:	Language materials, manuscript : Monograph/item
Title/Author:	Algorithmic Fairness in Graph Machine Learning :/
Reminder of title:	Explanation, Optimization, and Certification.
Author:	Dong, Yushun.
Description:	1 online resource (152 pages)
Notes:	Source: Dissertations Abstracts International, Volume: 86-01, Section: B.
Contained By:	Dissertations Abstracts International86-01B.
Subject:	Computer science. -
Online resource:	click for full text (PQDT)
ISBN:	9798383211731

Algorithmic Fairness in Graph Machine Learning : = Explanation, Optimization, and Certification.
Dong, Yushun.

Algorithmic Fairness in Graph Machine Learning :Explanation, Optimization, and Certification. - 1 online resource (152 pages)

Source: Dissertations Abstracts International, Volume: 86-01, Section: B.

Thesis (Ph.D.)--University of Virginia, 2024.

Includes bibliographical references

Network data is ubiquitous across diverse domains such as credit scoring, social networking, recommendation systems, and medical diagnosis. In this landscape, graph machine learning algorithms, e.g., Graph Neural Networks (GNNs), have emerged as powerful tools for modeling such data and performing predictive tasks. However, despite the effectiveness of existing graph machine learning algorithms, they usually bear the problem of exhibiting bias in the prediction results. This is particularly alarming under high-stakes decision-making contexts, since these algorithms could play pivotal roles and influence life-altering choices for involved individuals. Consequently, it becomes paramount to delve into the root causes of such biases and improve the level of fairness, thereby enhancing the trustworthiness of these algorithms.Nevertheless, addressing such a problem is non-trivial, and a plethora of unresolved questions loom large. For example, why does bias arise in graph machine learning? How to quantitatively measure the exhibited bias under different notions? How can we mitigate these biases, ensuring fair outcomes when these algorithms guide critical decisions? How to robustify the fairness level of graph machine learning algorithms against potential malicious attacks? How to remove the sensitive information that may lead to bias when it has been encoded in the graph machine learning algorithms in the training stage? It is necessary to properly answer these questions to ensure the trustworthiness of graph machine learning algorithms deployed in real-world applications. However, despite the significance of the algorithmic bias issue, the corresponding study remains at an early stage.To answer the questions above, my Ph.D. dissertation mainly contributes to the advancement of graph machine learning through a fairness lens. Specifically, this dissertation mainly focuses on three research themes, including algorithmic fairness explanation in graph machine learning, algorithmic fairness optimization in graph machine learning, and fairness certification in graph machine learning. In the first theme, we present qualitative and quantitative analysis to understand why bias arises and where the bias comes from in graph machine learning. In the second theme, we present principled frameworks and strategies to mitigate the bias exhibited in graph machine learning algorithms, where multiple fairness notions are taken into consideration. In the third theme, we propose theoretical certification on graph machine learning algorithms to achieve defense on the level of fairness and removal of sensitive information that may bring bias from the model to achieve protection for privacy.With these research themes, this dissertation aims to enhance the understanding of fair decision-making in the realm of graph machine learning and paves the way for future advances.

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

Mode of access: World Wide Web

ISBN: 9798383211731Subjects--Topical Terms:

573171
Computer science.
Subjects--Index Terms:

Algorithmic fairnessIndex Terms--Genre/Form:

554714
Electronic books.

Algorithmic Fairness in Graph Machine Learning : = Explanation, Optimization, and Certification.
LDR:04333ntm a22004097 4500 001 1149908
005 20241022110547.5
006 m o d
007 cr bn ---uuuuu
008 250605s2024 xx obm 000 0 eng d
020 $a 9798383211731
035 $a (MiAaPQ)AAI31333470
035 $a AAI31333470
040 $a MiAaPQ $b eng $c MiAaPQ $d NTU
100 1 $a Dong, Yushun. $3 1476282
245 1 0 $a Algorithmic Fairness in Graph Machine Learning : $b Explanation, Optimization, and Certification.
264 0 $c 2024
300 $a 1 online resource (152 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: 86-01, Section: B.
500 $a Advisor: Li, Jundong;Shen, Cong.
502 $a Thesis (Ph.D.)--University of Virginia, 2024.
504 $a Includes bibliographical references
520 $a Network data is ubiquitous across diverse domains such as credit scoring, social networking, recommendation systems, and medical diagnosis. In this landscape, graph machine learning algorithms, e.g., Graph Neural Networks (GNNs), have emerged as powerful tools for modeling such data and performing predictive tasks. However, despite the effectiveness of existing graph machine learning algorithms, they usually bear the problem of exhibiting bias in the prediction results. This is particularly alarming under high-stakes decision-making contexts, since these algorithms could play pivotal roles and influence life-altering choices for involved individuals. Consequently, it becomes paramount to delve into the root causes of such biases and improve the level of fairness, thereby enhancing the trustworthiness of these algorithms.Nevertheless, addressing such a problem is non-trivial, and a plethora of unresolved questions loom large. For example, why does bias arise in graph machine learning? How to quantitatively measure the exhibited bias under different notions? How can we mitigate these biases, ensuring fair outcomes when these algorithms guide critical decisions? How to robustify the fairness level of graph machine learning algorithms against potential malicious attacks? How to remove the sensitive information that may lead to bias when it has been encoded in the graph machine learning algorithms in the training stage? It is necessary to properly answer these questions to ensure the trustworthiness of graph machine learning algorithms deployed in real-world applications. However, despite the significance of the algorithmic bias issue, the corresponding study remains at an early stage.To answer the questions above, my Ph.D. dissertation mainly contributes to the advancement of graph machine learning through a fairness lens. Specifically, this dissertation mainly focuses on three research themes, including algorithmic fairness explanation in graph machine learning, algorithmic fairness optimization in graph machine learning, and fairness certification in graph machine learning. In the first theme, we present qualitative and quantitative analysis to understand why bias arises and where the bias comes from in graph machine learning. In the second theme, we present principled frameworks and strategies to mitigate the bias exhibited in graph machine learning algorithms, where multiple fairness notions are taken into consideration. In the third theme, we propose theoretical certification on graph machine learning algorithms to achieve defense on the level of fairness and removal of sensitive information that may bring bias from the model to achieve protection for privacy.With these research themes, this dissertation aims to enhance the understanding of fair decision-making in the realm of graph machine learning and paves the way for future advances.
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 Ethics. $3 555769
650 4 $a Information science. $3 561178
653 $a Algorithmic fairness
653 $a Graph machine learning
653 $a Graph Neural Networks
653 $a Decision-making
653 $a Real-world applications
655 7 $a Electronic books. $2 local $3 554714
690 $a 0984
690 $a 0394
690 $a 0723
690 $a 0800
710 2 $a ProQuest Information and Learning Co. $3 1178819
710 2 $a University of Virginia. $b Electrical Engineering. $3 1476283
773 0 $t Dissertations Abstracts International $g 86-01B.
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=31333470 $z click for full text (PQDT)