A Parallel Supercomputer Implementation of a Biological Inspired Neural Network and its use for Pattern Recognition

Abstract

Abstract. A parallel implementation of a large spiking neural network is proposed and evaluated. The neural network implements the binding by synchrony process using the Oscillatory Dynamic Link Matcher (ODLM). Scalability, speed and performance are compared for 2 implementations: Message Passing Interface (MPI) and Compute Unified Device Architecture (CUDA) running on clusters of multicore supercomputers and NVIDIA graphical processing units respectively. A global spiking list that represents at each instant the state of the neural network is described. This list indexes each neuron that fires during the current simulation time so that the influence of their spikes are simultaneously processed on all computing units. Our implementation shows a good scalability for very large networks. A complex and large spiking neural network has been implemented in parallel with success, thus paving the road towards real-life applications based on networks of spiking neurons. MPI offers a better scalability than CUDA, while the CUDA implementation on a GeForce GTX 285 gives the best cost to performance ratio. When running the neural network on the GTX 285, the processing speed is comparable to the MPI implementation on RQCHP's Mammouth parallel with 64 notes (128 cores).

Neural visual objects enhancement for sensorial substitution from vision to audition

“…Then the object retrieval and location should be easier. After that enhancement, the dynamic link matching system [22] could be used to more precisely locate objects in images.…”

An efficient automated parameter tuning framework for spiking neural networks

“…GPU acceleration has been used to increase the throughput of EAs (Maitre et al, 2009 ), simulate neural field models of the primary visual cortex V1 (Baladron et al, 2012 ), and search parameter spaces in bio-inspired object-recognition models (Pinto et al, 2009 ). In addition to these applications, a number of research groups in the computational neuroscience community (Brette and Goodman, 2012 ) have developed and implemented parallel implementations of SNNs on GPUs (Bernhard and Keriven, 2006 ; Fidjeland et al, 2009 ; Nageswaran et al, 2009b ; Bhuiyan et al, 2010 ; Han and Taha, 2010 ; Hoffmann et al, 2010 ; Yudanov et al, 2010 ; Ahmadi and Soleimani, 2011 ; Nowotny, 2011 ; Thibeault et al, 2011 ; de Ladurantaye et al, 2012 ; Mirsu et al, 2012 ; Pallipuram et al, 2012 ). GPU-driven SNN simulators have already been used in SNN models of the basal forebrain (Avery et al, 2012 ), the basal ganglia (Igarashi et al, 2011 ), the cerebellum (Yamazaki and Igarashi, 2013 ), and the olfactory system (Nowotny, 2010 ).…”