Angelo Carlone
Enhanced synaptic connectivity of spiking neural multiprocessor.
Rel. Maurizio Martina, Jordi Madrenas Boadas. Politecnico di Torino, Corso di laurea magistrale in Ingegneria Elettronica (Electronic Engineering), 2026
|
|
PDF (Tesi_di_laurea)
- Tesi
Accesso limitato a: Solo utenti staff fino al 2 Ottobre 2027 (data di embargo). Licenza: Creative Commons Attribution Non-commercial No Derivatives. Download (15MB) |
Abstract
Neuromorphic hardware distributes computation and storage across many simple, parallel processing elements, trading the generality of a conventional processor for efficiency and real-time operation. HEENS (Hardware Emulator of Evolving Neural Spiking Systems) is a programmable emulator of this kind, organised as a single-instruction, multiple-data array of processing elements. Synaptic connectivity is represented in a memory local to each processing element and can be reconfigured at run time, which gives the platform considerable flexibility. This flexibility has a limit, however: the connection memory has a fixed capacity. An application in which every neuron of one group connects to every neuron of the next, such as a synfire chain, can therefore require more synapses per receiving neuron than this memory can hold, as the chain grows in length and group size.
This thesis introduces a connectivity mechanism that supplies the additional synapses without enlarging the connection memory: possible presynaptic neurons are addressed directly through their position in the processing-element array, rather than through a stored, per-synapse entry
Relatori
Anno Accademico
Tipo di pubblicazione
Numero di pagine
Corso di laurea
Classe di laurea
Ente in cotutela
Aziende collaboratrici
URI
![]() |
Modifica (riservato agli operatori) |
