Réseaux de neurones sur silicium : une approche mixte, analogique / numérique, pour l'étude des phénomènes d'adaptation, d'apprentissage et de plasticité

Abstract : This work addresses the integration issues for neurominetic integrated circuits de-dicated to computationnal neuroscience, in the context of adaptive neural networks. Those circuits implement a biologically realistic model. This model is continuously solved in real time. the network adaptation is also computed in real time by digital means using sofware and or hardware material. The software part of the system is handled by a computer, interfaced to the system through the PCI bus, whereas the hardware part uses FPGAS. Three generations of the simulation system are presented, followed by a critical analysis on their performance as analog neural network simulators.
Complete list of metadatas

https://tel.archives-ouvertes.fr/tel-00181353
Contributor : Sylvain Saighi <>
Submitted on : Tuesday, October 23, 2007 - 3:28:44 PM
Last modification on : Thursday, January 11, 2018 - 6:21:07 AM
Long-term archiving on: Sunday, April 11, 2010 - 11:55:27 PM

Identifiers

  • HAL Id : tel-00181353, version 1

Citation

Yannick Bornat. Réseaux de neurones sur silicium : une approche mixte, analogique / numérique, pour l'étude des phénomènes d'adaptation, d'apprentissage et de plasticité. Micro et nanotechnologies/Microélectronique. Université Sciences et Technologies - Bordeaux I, 2006. Français. ⟨tel-00181353⟩

Share

Metrics

Record views

453

Files downloads

974