Representation Learning on Relational Data
MCML Authors
Evgeny Faerman
Dr.
* Former Member
Abstract
Evgeny Faerman
Dr.
* Former Member
Abstract
This thesis introduces methods that leverage relational information to address various problems in machine learning, such as node classification, graph matching, and argument mining. It explores unsupervised and semi-supervised approaches for node classification, graph alignment for geographical maps and knowledge graphs, and proposes a novel method for identifying and searching arguments in peer reviews. Additionally, it presents a subspace clustering method that uses relationships to improve clustering performance on large datasets. (Shortened.)
BibTeXKey: Fae21