Higher order knowledge mining for humanistic data-generated graphs
Date Issued
November 11, 2024
Type
Διδακτορική Διατριβή
Abstract
Humanistic data not only are important by definition, but also are gradually becoming the prevalent data category in contemporary knowledge-based civilizations. They have a plethora of important properties including a potentially enormous size, increasingly quicker generation rates, and maybe corrupted or incomplete entries. Despite these challenges mining these data is bound to provide crucial knowledge across numerous aspects of everyday life smart cities, digital health, and electronic financial transactions. Thus, algorithms handling humanistic linked oriented data should take these properties into consideration and, wherever possible, turn them to their advantage. To this end machine learning is critical. Three neural network architectures, namely tensor stack networks, self organizing maps, and graph neural networks have been applied respectively to the assessment of graph resiliency, the clustering of fMRI images, and the discovery of affective communities in Twitter graphs.
Subjects