neker
Title: |
neker |
DNr: |
LiU-2019-27 |
Project Type: |
LiU Compute |
Principal Investigator: |
Magnus Sethson <magnus.sethson@liu.se> |
Affiliation: |
Linköpings universitet |
Duration: |
2019-06-07 – 2023-07-01 |
Classification: |
20207 |
Keywords: |
|
Abstract
neker is a first attempt for a spiking neuron network computational engine aiming for deep learning tests with a highly efficient memory footprint. Its unique feature is to project an unstructured net onto a structured computational core. This is accomplished by mixing sparse and dense matrix operations using C functions primarily from the GSL project. (neker means "sädeskärve" in old Swedish language and refers to its structured look). The main application is for evaluating condition monitoring strategies using a large set of measurements on a pneumatic test stand in our laboratory here at LiU. The project goal is to develop a small and efficient spiking neural network engine that can possible run on embedded systems. Training need super computer power though.