Giovanni Donghi
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Giovanni Donghi has a total of 20 co-authors. Here are the top co-authors by number of shared publications:
Luca Pasa
University of Padua
4 shared publicationsNicolò Navarin
University of Padua
4 shared publicationsAlessandro Sperduti
University of Padua
3 shared publications- Nicolò Navarin
University of Padova, Padua, Italy
3 shared publications - Alessandro Sperduti
University of Padua
2 shared publications Cesare Alippi
Università della Svizzera Italiana
2 shared publications
Latest Publications
- 22 September 2026
- Published by Springer Nature
- p. 458-470
How Streaming LDA Questions our Progress in Continual Learning with Pretrained Models
- 22 September 2026
- Published by Springer Nature
- p. 458-470
How Streaming LDA Questions our Progress in Continual Learning with Pretrained Models
- Open Access
- 15 September 2026
- Published by Springer Nature in Machine Learning
- Vol. 115 (9) , 220
The Unreasonable Effectiveness of Randomized Representations in Online Continual Graph Learning
Catastrophic forgetting is one of the main obstacles for Online Continual Graph Learning (OCGL), where nodes arrive one by one, distribution drifts may occur at any time and offline training on task-specific subgraphs is not feasible. In this work, we explore a surprisingly simple yet highly effective approach for OCGL: we use a fixed, randomly initialized encoder to generate robust and expressive node embeddings by aggregating neighborhood information, training online only a lightweight classifier. By freezing the encoder, we eliminate drifts of the representation parameters, a key source of forgetting, obtaining embeddings that are both expressive and stable. When evaluated across several OCGL benchmarks, despite its simplicity and lack of memory buffer, this approach yields consistent gains over state-of-the-art methods, with surprising improvements of up to 30% and performance often approaching that of the joint offline-training upper bound. These results suggest that in OCGL, catastrophic forgetting can be minimized without complex replay or regularization by embracing architectural simplicity and stability.
- 15 September 2026
- Published by Springer Nature in Machine Learning
- Vol. 115 (9) , 220
Open AccessThe Unreasonable Effectiveness of Randomized Representations in Online Continual Graph Learning
Catastrophic forgetting is one of the main obstacles for Online Continual Graph Learning (OCGL), where nodes arrive one by one, distribution drifts may occur at any time and offline training on task-specific subgraphs is not feasible. In this work, we explore a surprisingly simple yet highly effective approach for OCGL: we use a fixed, randomly initialized encoder to generate robust and expressive node embeddings by aggregating neighborhood information, training online only a lightweight classifier. By freezing the encoder, we eliminate drifts of the representation parameters, a key source of forgetting, obtaining embeddings that are both expressive and stable. When evaluated across several OCGL benchmarks, despite its simplicity and lack of memory buffer, this approach yields consistent gains over state-of-the-art methods, with surprising improvements of up to 30% and performance often approaching that of the joint offline-training upper bound. These results suggest that in OCGL, catastrophic forgetting can be minimized without complex replay or regularization by embracing architectural simplicity and stability.
- 15 September 2026
- Published by Springer Nature
- p. 172-184
From Concepts to Relations: Generative Models as Relational Associative Memories
- 15 September 2026
- Published by Springer Nature
- p. 172-184
From Concepts to Relations: Generative Models as Relational Associative Memories
- Open Access
The Unreasonable Effectiveness of Randomized Representations in Online Continual Graph Learning
Catastrophic forgetting is one of the main obstacles for Online Continual Graph Learning (OCGL), where nodes arrive one by one, distribution drifts may occur at any time and offline training on task-specific subgraphs is not feasible. In this work, we explore a surprisingly simple yet highly effective approach for OCGL: we use a fixed, randomly initialized encoder to generate robust and expressive node embeddings by aggregating neighborhood information, training online only a lightweight classifier. By freezing the encoder, we eliminate drifts of the representation parameters, a key source of forgetting, obtaining embeddings that are both expressive and stable. When evaluated across several OCGL benchmarks, despite its simplicity and lack of memory buffer, this approach yields consistent gains over state-of-the-art methods, with surprising improvements of up to 30% and performance often approaching that of the joint offline-training upper bound. These results suggest that in OCGL, catastrophic forgetting can be minimized without complex replay or regularization by embracing architectural simplicity and stability.
Open AccessThe Unreasonable Effectiveness of Randomized Representations in Online Continual Graph Learning
Catastrophic forgetting is one of the main obstacles for Online Continual Graph Learning (OCGL), where nodes arrive one by one, distribution drifts may occur at any time and offline training on task-specific subgraphs is not feasible. In this work, we explore a surprisingly simple yet highly effective approach for OCGL: we use a fixed, randomly initialized encoder to generate robust and expressive node embeddings by aggregating neighborhood information, training online only a lightweight classifier. By freezing the encoder, we eliminate drifts of the representation parameters, a key source of forgetting, obtaining embeddings that are both expressive and stable. When evaluated across several OCGL benchmarks, despite its simplicity and lack of memory buffer, this approach yields consistent gains over state-of-the-art methods, with surprising improvements of up to 30% and performance often approaching that of the joint offline-training upper bound. These results suggest that in OCGL, catastrophic forgetting can be minimized without complex replay or regularization by embracing architectural simplicity and stability.
- Open Access
- 24 August 2025
- Published by Elsevier in Neural Networks
- Vol. 193, 108029
On the application of neural networks for structured domains to fMRI data
- 24 August 2025
- Published by Elsevier in Neural Networks
- Vol. 193, 108029
Open AccessOn the application of neural networks for structured domains to fMRI data
- 17 September 2024
- Published by Springer Nature
- p. 373-383
Relative Local Signal Strength: The Impact of Normalization on the Analysis of Neuroimaging Data with Deep Learning
- 17 September 2024
- Published by Springer Nature
- p. 373-383
Relative Local Signal Strength: The Impact of Normalization on the Analysis of Neuroimaging Data with Deep Learning
- Open Access
- 7 August 2024
- Published by Elsevier in Neurocomputing
- Vol. 606
Investigating over-parameterized randomized graph networks
- 7 August 2024
- Published by Elsevier in Neurocomputing
- Vol. 606
Open AccessInvestigating over-parameterized randomized graph networks