ProgramsView all programs
From the news
- Chancellor Amma Addresses the Parliament of World’s Religions
- Amrita Students Qualify for the European Mars Rover Challenge
Publication Type : Conference Paper
Thematic Areas : Learning-Technologies, Medical Sciences, Biotech
Publisher : Proceedings of the 2014 International Conference on Interdisciplinary Advances in Applied Computing, ACM, New York, NY, USA.
Source : Proceedings of the 2014 International Conference on Interdisciplinary Advances in Applied Computing, ACM, New York, NY, USA (2014)
ISBN : 9781450329088
Keywords : Cerebellum Granule Cell, Computational neuroscience, MPI, parallel computing
Campus : Amritapuri
School : School of Biotechnology
Center : Amrita Mind Brain Center, Biotechnology, Computational Neuroscience and Neurophysiology
Department : Computational Neuroscience Laboratory, biotechnology
Year : 2014
Abstract : Rapid progress in biophysical neural network modelling has been observed in the last years as a focus within computational neuroscience. Detailed multi-compartmental neuron models that were built to simulate physiological aspects of cerebellar neurons and microcircuits involve hundreds of equations. Simulating several hundreds of neurons is computationally expensive. Storage of data and run-time evaluations also prove to be major challenges in this kind of scenario, which limits the researchers.In this paper, we use detailed models of neurons reconstructing the biophysics of cable properties and action ion channel models to generate a neural micro circuitry of cerebellum input layer. We report the process of adapting and profiling a parallel, MPI-based version of the network model on NEURON for large-scale simulations. Using multi-split and distributed approaches, our model was parallelized on multi-core, multi-processor systems. A spatio-temporal activation pattern, called the centre-surround was elicited in the model validating the biophysical role of synaptic inhibition modulating excitatory activation, usually observed during sensory or tactile stimulation. Performance tests were carried out on two heterogeneous computing clusters. We see a significant reduction of computational cost in terms of power and time while simulating parallelized code although the most apt method depended on network size and nature of synaptic connections. We find 'embarrassingly parallel' method augmented efficiency in terms of processor core usage and also decreased simulation time.
Cite this Research Publication : A. Yoosef, Harilal Parasuram, Chaitanya Medini, Sergio Solinas, Egidio D'Angelo, Dr. Bipin G. Nair, and Dr. Shyam Diwakar, “Parallelization of a Computational Model of a Biophysical Neuronal Circuitry of Rat Cerebellum”, in Proceedings of International Conference on Interdisciplinary Advances in Applied Computing, Amrita University, Coimbatore, India, October 10-11, 2014.