Mascione, D. and Cristoforetti, M. and Di Luca, A. and Follega, F. M. and Iuppa, R. (2023) Innovative deep neural networks resizing for FPGA implementation in future collider experiments. Il nuovo cimento C, 46 (4). pp. 1-4. ISSN 1826-9885
|
Text
ncc12667.pdf - Published Version Download (184kB) | Preview |
Abstract
Future collider experiments are expected to produce thousands of ExaBytes of data per year, and it will become impractical to maintain old trigger schemes to preserve data for offline analysis. Deep Neural Networks offer a great opportunity to take fast and accurate decisions, but it is essential to employ Deep Learning algorithms online at the early stage of event selection, taking advantage of FPGAs. In order to optimize Deep Neural Networks under size limits to accommodate resources of FPGAs, we developed a novel pruning technique. The proposed method can reduce the overall sizes of neural networks by pruning unnecessary nodes, with the network final dimensions determined by the user.
| Item Type: | Article |
|---|---|
| Subjects: | 500 Scienze naturali e Matematica > 530 Fisica |
| Depositing User: | Marina Spanti |
| Date Deposited: | 02 Feb 2024 10:54 |
| Last Modified: | 02 Feb 2024 10:54 |
| URI: | http://eprints.bice.rm.cnr.it/id/eprint/22670 |
Actions (login required)
 |
View Item |
