Rosario Napoli
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Rosario Napoli has a total of 9 co-authors. Here are the top co-authors by number of shared publications:
- Antonio Celesti
University of Messina
7 shared publications - Maria Fazio
University of Messina
7 shared publications - Massimo Villari
University of Messina
7 shared publications - Gabriele Morabito
Dipartimento di Dipartimento di Scienze Biomediche, Odontoiatriche e delle Immagini Morfologiche e Funzionali, Università degli Studi di Messina, Messina, Sicily, Italy
2 shared publications - Giovanni Lonia
University of Messina
2 shared publications Antonio Celesti
1 shared publication
Latest Publications
- Open Access
- 6 July 2026
- Published by Elsevier in SSRN Electronic Journal
Bi-View Fusion: Combining Transductive and Inductive Embeddings for Knowledge Graph’s Nodes Classification
Node classification on knowledge graphs (KGs) is challenged by sparse node features and implicit semantics encoded in graph topology rather than explicit attributes, limiting the applicability of conventional graph machine learning (GML) approaches. This paper proposes Bi-View, a hybrid embedding strategy that addresses these limitations by enriching node features through the combination of Node2Vec structural embeddings and graph-theoretic centrality metrics used as auxiliary input to Graph Sample and aggreGatE(GraphSAGE). The resulting structural and neighbourhood-aggregated representations are subsequently fused by a supervised multilayer perceptron (FusionNet), trained jointly with cross-entropy and online hard triplet loss to produce discriminative embeddings that capture both topological and task-specific semantic information. The proposed approach is evaluated on five heterogeneous KG datasets across four embedding dimensionalities using multi-axis evaluation framework comprising intrinsic cluster separability (Silhouette Score), cluster-to-class alignment (Adjusted Rand Index), and downstream macro F1-score computed over 20 independent repetitions. In experiments, we compared both Bi-View and Bi-View Pure (i.e., an ablated variant) with the major strategies available in the literature. Experiments indicate that Bi-View achieves competitive node classification performance among methods that do not condition on explicit relation-type labels, and the multi-axis evaluation reveals non-trivial discrepancies between intrinsic embedding geometry and downstream discriminative utility.
- 6 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessBi-View Fusion: Combining Transductive and Inductive Embeddings for Knowledge Graph’s Nodes Classification
Node classification on knowledge graphs (KGs) is challenged by sparse node features and implicit semantics encoded in graph topology rather than explicit attributes, limiting the applicability of conventional graph machine learning (GML) approaches. This paper proposes Bi-View, a hybrid embedding strategy that addresses these limitations by enriching node features through the combination of Node2Vec structural embeddings and graph-theoretic centrality metrics used as auxiliary input to Graph Sample and aggreGatE(GraphSAGE). The resulting structural and neighbourhood-aggregated representations are subsequently fused by a supervised multilayer perceptron (FusionNet), trained jointly with cross-entropy and online hard triplet loss to produce discriminative embeddings that capture both topological and task-specific semantic information. The proposed approach is evaluated on five heterogeneous KG datasets across four embedding dimensionalities using multi-axis evaluation framework comprising intrinsic cluster separability (Silhouette Score), cluster-to-class alignment (Adjusted Rand Index), and downstream macro F1-score computed over 20 independent repetitions. In experiments, we compared both Bi-View and Bi-View Pure (i.e., an ablated variant) with the major strategies available in the literature. Experiments indicate that Bi-View achieves competitive node classification performance among methods that do not condition on explicit relation-type labels, and the multi-axis evaluation reveals non-trivial discrepancies between intrinsic embedding geometry and downstream discriminative utility.
- Open Access
- 31 December 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-10
Knowledge Graphs-Driven Intelligence for Distributed Decision Systems
- 31 December 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-10
Open AccessKnowledge Graphs-Driven Intelligence for Distributed Decision Systems
- Open Access
- 24 December 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-10
How to evaluate NoSQL Database Paradigms for Knowledge Graph Processing
- 24 December 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-10
Open AccessHow to evaluate NoSQL Database Paradigms for Knowledge Graph Processing
- Open Access
Improving Graph Embeddings in Machine Learning Using Knowledge Completion with Validation in a Case Study on COVID-19 Spread
The rise of graph-structured data has driven major advances in Graph Machine Learning (GML), where graph embeddings (GEs) map features from Knowledge Graphs (KGs) into vector spaces, enabling tasks like node classification and link prediction. However, since GEs are derived from explicit topology and features, they may miss crucial implicit knowledge hidden in seemingly sparse datasets, affecting graph structure and their representation. We propose a GML pipeline that integrates a Knowledge Completion (KC) phase to uncover latent dataset semantics before embedding generation. Focusing on transitive relations, we model hidden connections with decay-based inference functions, reshaping graph topology, with consequences on embedding dynamics and aggregation processes in GraphSAGE and Node2Vec. Experiments show that our GML pipeline significantly alters the embedding space geometry, demonstrating that its introduction is not just a simple enrichment but a transformative step that redefines graph representation quality.
Open AccessImproving Graph Embeddings in Machine Learning Using Knowledge Completion with Validation in a Case Study on COVID-19 Spread
The rise of graph-structured data has driven major advances in Graph Machine Learning (GML), where graph embeddings (GEs) map features from Knowledge Graphs (KGs) into vector spaces, enabling tasks like node classification and link prediction. However, since GEs are derived from explicit topology and features, they may miss crucial implicit knowledge hidden in seemingly sparse datasets, affecting graph structure and their representation. We propose a GML pipeline that integrates a Knowledge Completion (KC) phase to uncover latent dataset semantics before embedding generation. Focusing on transitive relations, we model hidden connections with decay-based inference functions, reshaping graph topology, with consequences on embedding dynamics and aggregation processes in GraphSAGE and Node2Vec. Experiments show that our GML pipeline significantly alters the embedding space geometry, demonstrating that its introduction is not just a simple enrichment but a transformative step that redefines graph representation quality.
- 13 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 300-307
Improving Graph Embeddings in Machine Learning Using Knowledge Completion with Validation in a Case Study on COVID-19 Spread
The rise of graph-structured data has driven major advances in Graph Machine Learning (GML), where graph embeddings (GEs) map features from Knowledge Graphs (KGs) into vector spaces, enabling tasks like node classification and link prediction. However, since GEs are derived from explicit topology and features, they may miss crucial implicit knowledge hidden in seemingly sparse datasets, affecting graph structures and their representation. We propose a GML pipeline that integrates a Knowledge Completion (KC) phase to uncover latent dataset semantics before embedding generation. Focusing on transitive relations, we model hidden connections with decay-based inference functions, reshaping graph topology, with consequences on embedding dynamics and aggregation processes in GraphSAGE and Node2Vec. Experiments show that our GML pipeline significantly alters the embedding space geometry, demonstrating that its introduction is not just a simple enrichment but a transformative step that redefines graph representation quality.
- 13 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 300-307
Improving Graph Embeddings in Machine Learning Using Knowledge Completion with Validation in a Case Study on COVID-19 Spread
The rise of graph-structured data has driven major advances in Graph Machine Learning (GML), where graph embeddings (GEs) map features from Knowledge Graphs (KGs) into vector spaces, enabling tasks like node classification and link prediction. However, since GEs are derived from explicit topology and features, they may miss crucial implicit knowledge hidden in seemingly sparse datasets, affecting graph structures and their representation. We propose a GML pipeline that integrates a Knowledge Completion (KC) phase to uncover latent dataset semantics before embedding generation. Focusing on transitive relations, we model hidden connections with decay-based inference functions, reshaping graph topology, with consequences on embedding dynamics and aggregation processes in GraphSAGE and Node2Vec. Experiments show that our GML pipeline significantly alters the embedding space geometry, demonstrating that its introduction is not just a simple enrichment but a transformative step that redefines graph representation quality.
- 21 February 2025
- Published by Springer Nature
- p. 57-64
A Bio-inspired Leader-Based Energy Management System for Drone Fleets
- 21 February 2025
- Published by Springer Nature
- p. 57-64
A Bio-inspired Leader-Based Energy Management System for Drone Fleets