SNAVA: A Generic Threshold-Based-SNN Emulation Solution

[ANGLÈS] Spiking Neural Networks has been the focus of research of the decade. It has been proved possible to mimic several mechanisms of the mammalian brain with SNN models. Several dedicated hardware, both analog and digital, have been built for SNN emulation in the past decade. In this work an ex...

Descripción completa

Detalles Bibliográficos
Autor: Tiruvendipura Achyutha Raghavan, Athul Sripad
Tipo de recurso: tesis de maestría
Fecha de publicación:2013
País:España
Institución:Universitat Politècnica de Catalunya (UPC)
Repositorio:UPCommons. Portal del coneixement obert de la UPC
Idioma:inglés
OAI Identifier:oai:upcommons.upc.edu:2099.1/19633
Acceso en línea:https://hdl.handle.net/2099.1/19633
Access Level:acceso abierto
Palabra clave:Neural networks (Computer science)
FPGA
Spiking Neural Networks
SIMD
Digital Logic Design
Algorithms and Architectures
Redes Neuronales Spiking
Diseño Digital
Algoritmos y arquitecturas
Xarxes neuronals (Informàtica)
Àrees temàtiques de la UPC::Enginyeria de la telecomunicació
Descripción
Sumario:[ANGLÈS] Spiking Neural Networks has been the focus of research of the decade. It has been proved possible to mimic several mechanisms of the mammalian brain with SNN models. Several dedicated hardware, both analog and digital, have been built for SNN emulation in the past decade. In this work an existing SIMD processor architecture called Ubichip has been analyzed, bottlenecks have been identified and improvements have been proposed. A new vector processing Architecture called SNAVA has been proposed with features of virtualization and real-time parameter monitoring. It has been implemented on a Xilinx KC705 kit. A proof of concept application has also been developed and experimental results have been presented.