Ponencia
ConvNets Experiments on SpiNNaker
Autor/es | Serrano Gotarredona, María Teresa
Linares Barranco, Bernabé Galluppi, Francesco Plana, L. Furber, S. |
Departamento | Universidad de Sevilla. Departamento de Arquitectura y Tecnología de Computadores |
Fecha de publicación | 2015 |
Fecha de depósito | 2020-10-20 |
Publicado en |
|
ISBN/ISSN | 978-1-4799-8391-9 0271-4302 |
Resumen | The SpiNNaker Hardware platform allows
emulating generic neural network topologies, where each neuronto-
neuron connection is defined by an independent synaptic
weight. Consequently, weight storage requires an ... The SpiNNaker Hardware platform allows emulating generic neural network topologies, where each neuronto- neuron connection is defined by an independent synaptic weight. Consequently, weight storage requires an important amount of memory in the case of generic neural network topologies. This is solved in SpiNNaker by encapsulating with each SpiNNaker chip (which includes 18 ARM cores) a 128MB DRAM chip within the same package. However, ConvNets (Convolutional Neural Network) posses "weight sharing" property, so that many neuron-to-neuron connections share the same weight value. Therefore, a very reduced amount of memory is required to define all synaptic weights, which can be stored on local SRAM DTCM (data-tightly-coupled-memory) at each ARM core. This way, DRAM can be used extensively to store traffic data for off-line analyses. We show an implementation of a 5- layer ConvNet for symbol recognition. Symbols are obtained with a DVS camera. Neurons in the ConvNet operate in an eventdriven fashion, and synapses operate instantly. With this approach it was possible to allocate up to 2048 neurons per ARM core, or equivalently 32k neurons per SpiNNaker chip. |
Agencias financiadoras | Ministerio de Educación, Cultura y Deporte (MECD). España Ministerio de Educación, Cultura y Deporte (MECD). España European Union (UE) Ministerio de Economía y Competitividad (MINECO). España Ministerio de Economía y Competitividad (MINECO). España Junta de Andalucía |
Identificador del proyecto | PRX12/00562
PRX12/00558 FP7-ICT-2013-FET-F-604102 PRI-PIMCHI-2011-0768 TEC2012-37868-C04-01 TIC-2010-6091 |
Cita | Serrano Gotarredona, M.T., Linares Barranco, B., Galluppi, F., Plana, L. y Furber, S. (2015). ConvNets Experiments on SpiNNaker. En ISCAS 2015: IEEE International Symposium on Circuits and Systems (2405-2408), Lisboa, Portugal: IEEE Computer Society. |
Ficheros | Tamaño | Formato | Ver | Descripción |
---|---|---|---|---|
ConvNets experiments on SpiNNa ... | 1.454Mb | [PDF] | Ver/ | |