Utilize este identificador para referenciar este registo:
http://hdl.handle.net/10071/29853
Autoria: | Napoli, O. O. Almeida, A. M. de. Dias, J. M. S. Rosário, L. B. Borin, E. Breternitz Jr, M. |
Data: | 2023 |
Título próprio: | Efficient knowledge aggregation methods for weightless neural networks |
Título e volume do livro: | Proceedings of the 31th European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning (ESANN 2023) |
Paginação: | 369 - 374 |
Título do evento: | 31th European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning (ESANN 2023) |
Referência bibliográfica: | Napoli, O. O., Almeida, A. M. de., Dias, J. M. S., Rosário, L. B., Borin, E., & Breternitz Jr, M. (2023). Efficient knowledge aggregation methods for weightless neural networks. In Proceedings of the 31th European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning (ESANN 2023) (pp. 369-374). ESANN. https://doi.org/10.14428/esann/2023.ES2023-123 |
ISBN: | 978-2-87587-088-9 |
DOI (Digital Object Identifier): | 10.14428/esann/2023.ES2023-123 |
Resumo: | Weightless Neural Networks (WNN) are good candidates for Federated Learning scenarios due to their robustness and computational lightness. In this work, we show that it is possible to aggregate the knowledge of multiple WNNs using more compact data structures, such as Bloom Filters, to reduce the amount of data transferred between devices. Finally, we explore variations of Bloom Filters and found that a particular data-structure, the Count-Min Sketch (CMS), is a good candidate for aggregation. Costing at most 3% of accuracy, CMS can be up to 3x smaller when compared to previous approaches, specially for large datasets. |
Arbitragem científica: | yes |
Acesso: | Acesso Aberto |
Aparece nas coleções: | ISTAR-CRI - Comunicações a conferências internacionais |
Ficheiros deste registo:
Ficheiro | Tamanho | Formato | |
---|---|---|---|
conferenceobject_97785.pdf | 1,69 MB | Adobe PDF | Ver/Abrir |
Todos os registos no repositório estão protegidos por leis de copyright, com todos os direitos reservados.