We introduce an architecture based on deep hierarchical decompositions to learn effective representations of large graphs. Our framework extends classic R-decompositions used in kernel methods, enabling nested part-of-part relations. Unlike recursive neural networks, which unroll a template on input graphs directly, we unroll a neural network template over the decomposition hierarchy, allowing us to deal with the high degree variability that typically characterize social network graphs. Deep hierarchical decompositions are also amenable to domain compression, a technique that reduces both space and time complexity by exploiting symmetries. We show empirically that our approach is able to outperform current state-of-the-art graph classification methods on large social network datasets, while at the same time being competitive on small chemobiological benchmark datasets.

Shift aggregate extract networks / Orsini, Francesco*; Baracchi, Daniele; Frasconi, Paolo. - In: FRONTIERS IN ROBOTICS AND AI. - ISSN 2296-9144. - ELETTRONICO. - 5:(2018), pp. 0-0. [10.3389/frobt.2018.00042]

Shift aggregate extract networks

Orsini, Francesco
;
Baracchi, Daniele;Frasconi, Paolo
2018

Abstract

We introduce an architecture based on deep hierarchical decompositions to learn effective representations of large graphs. Our framework extends classic R-decompositions used in kernel methods, enabling nested part-of-part relations. Unlike recursive neural networks, which unroll a template on input graphs directly, we unroll a neural network template over the decomposition hierarchy, allowing us to deal with the high degree variability that typically characterize social network graphs. Deep hierarchical decompositions are also amenable to domain compression, a technique that reduces both space and time complexity by exploiting symmetries. We show empirically that our approach is able to outperform current state-of-the-art graph classification methods on large social network datasets, while at the same time being competitive on small chemobiological benchmark datasets.
2018
5
0
0
Orsini, Francesco*; Baracchi, Daniele; Frasconi, Paolo
File in questo prodotto:
File Dimensione Formato  
frobt-05-00042.pdf

accesso aperto

Descrizione: Articolo principale
Tipologia: Pdf editoriale (Version of record)
Licenza: Creative commons
Dimensione 3.49 MB
Formato Adobe PDF
3.49 MB Adobe PDF

I documenti in FLORE sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificatore per citare o creare un link a questa risorsa: https://hdl.handle.net/2158/1150949
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 12
  • ???jsp.display-item.citation.isi??? 9
social impact