Pattern recognition in the graphs domain gained a lot of attention in the last two decades, since graphs are able to describe relationships (edges) between atomic entities (nodes) which can further be equipped with attributes encoding meaningful information. In this work, we investigate a novel graph embedding procedure based on the Granular Computing paradigm. Conversely to recently-developed techniques, we propose a stratified procedure for extracting suitable information granules (namely, frequent and/or meaningful subgraphs) in a class-aware fashion; that is, each class for the classification problem at hand is represented by the set of its own pivotal information granules. Computational results on several open-access datasets show performance improvements when considering also the ground-truth class labels in the information granulation procedure. Furthermore, since the granulation procedure is based on random walks, it is also very appealing in Big Data scenarios.
Towards a class-aware information granulation for graph embedding and classification / Baldini, L.; Martino, A.; Rizzi, A.. - (2021), pp. 263-290. (Intervento presentato al convegno 11th International Joint Conference on Computational Intelligence, IJCCI 2019 tenutosi a Vienna; Austria) [10.1007/978-3-030-70594-7_11].
Towards a class-aware information granulation for graph embedding and classification
Baldini L.
Primo
;Rizzi A.Ultimo
2021
Abstract
Pattern recognition in the graphs domain gained a lot of attention in the last two decades, since graphs are able to describe relationships (edges) between atomic entities (nodes) which can further be equipped with attributes encoding meaningful information. In this work, we investigate a novel graph embedding procedure based on the Granular Computing paradigm. Conversely to recently-developed techniques, we propose a stratified procedure for extracting suitable information granules (namely, frequent and/or meaningful subgraphs) in a class-aware fashion; that is, each class for the classification problem at hand is represented by the set of its own pivotal information granules. Computational results on several open-access datasets show performance improvements when considering also the ground-truth class labels in the information granulation procedure. Furthermore, since the granulation procedure is based on random walks, it is also very appealing in Big Data scenarios.File | Dimensione | Formato | |
---|---|---|---|
Baldini_Towards_2021.pdf
solo gestori archivio
Tipologia:
Versione editoriale (versione pubblicata con il layout dell'editore)
Licenza:
Tutti i diritti riservati (All rights reserved)
Dimensione
586.83 kB
Formato
Adobe PDF
|
586.83 kB | Adobe PDF | Contatta l'autore |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.