Development of load partitioning and mapping algorithms on neuromorphic SpiNNaker platform
External reference persons Gianvito Urgese, UniversitÓ di Manchester
Research Groups ELECTRONIC DESIGN AUTOMATION - EDA
Thesis type DEVELOPMENT
Description The thesis is in the context of the european project Human Brain Project (https://www.humanbrainproject.eu). It consists in the development of load mapping and allocation algorithms for a neuromorphic multicore platforms aimed at emulating the behaviour of the human brain. The platform, called SpiNNaker (http://apt.cs.manchester.ac.uk/projects/SpiNNaker/), emulates neurons and their communication through spikes using a densely interconnected board exploiting a Network-On-Chip. The purpose of the algorithm is to optimise the usage of the interconnection network exploiting an efficient mapping of neurons populations on the cores accounting for their communication characteristics.
Notes Thesis in collaboration with University of Manchester
Deadline 24/07/2015 PROPONI LA TUA CANDIDATURA