Stochastic Graphlet Embedding - Sorbonne Université
Article Dans Une Revue IEEE Transactions on Neural Networks and Learning Systems Année : 2019

Stochastic Graphlet Embedding

Résumé

Graph-based methods are known to be successful in many machine learning and pattern classification tasks. These methods consider semi-structured data as graphs where nodes correspond to primitives (parts, interest points, segments, etc.) and edges characterize the relationships between these primitives. However, these non-vectorial graph data cannot be straightforwardly plugged into off-the-shelf machine learning algorithms without a preliminary step of-explicit/implicit-graph vectorization and embedding. This embedding process should be resilient to intra-class graph variations while being highly discriminant. In this paper, we propose a novel high-order stochastic graphlet embedding (SGE) that maps graphs into vector spaces. Our main contribution includes a new stochastic search procedure that efficiently parses a given graph and extracts/samples unlimitedly high-order graphlets. We consider these graphlets, with increasing orders, to model local primitives as well as their increasingly complex interactions. In order to build our graph representation, we measure the distribution of these graphlets into a given graph, using particular hash functions that efficiently assign sampled graphlets into isomorphic sets with a very low probability of collision. When combined with maximum margin classifiers, these graphlet-based representations have positive impact on the performance of pattern comparison and recognition as corroborated through extensive experiments using standard benchmark databases.
Fichier principal
Vignette du fichier
1702.00156.pdf (1.49 Mo) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-02277646 , version 1 (03-09-2019)

Identifiants

Citer

Anjan Dutta, Hichem Sahbi. Stochastic Graphlet Embedding. IEEE Transactions on Neural Networks and Learning Systems, 2019, 30 (8), pp.2369-2382. ⟨10.1109/TNNLS.2018.2884700⟩. ⟨hal-02277646⟩
155 Consultations
101 Téléchargements

Altmetric

Partager

More