NEST 2: A Parallel Simulator for Large Neuronal Networks

Abstract

Event Abstract Back to Event NEST 2: A Parallel Simulator for Large Neuronal Networks Marc-Oliver Gewaltig1* 1 Honda Research Institute Europe GmbH, Germany NEST is a simulator for large heterogeneous networks of point neurons or neurons with few compartments. NEST supports parallel simulation on multi-core computers and computer clusters, with excellent scaling [1].NEST is best suited for models that focus on the dynamics, size, and structure of neural systems rather than on the detailed morphology of individual neurons. Examples are: Models of sensory processing in visual or auditory cortex. Models of network dynamics, e.g. in laminar cortical networks or random networks. Models of spike-synchronization in feed-forward networks such as synfire chains. Models of learning and plasticity. Here we demonstrate NEST 2 and invite visitors to try it interactively. The new Python-based user interface PyNEST makes it easier to learn and use NEST. Together with analysis packages like Scientific Python (www.scipy.org) and analysis tools from neuralensemble.org, users can now simulate networks and analyze results in one interactive Python session [2]. The new Topology module allows concise specification and efficient construction of large models with spatial structure, such as models of the visual system. A new multi-compartment model with soma, proximal and distal dendrite, provides a scaffold for users to implement their own multi-compartment models. Our new implementation of synaptic plasticity incorporates external factors such as dopamine modulation. Users can write new neuron models in NEST's simulation language SLI. The dynamic equations can be given in standard mathematical notation and NEST will compile them into efficient code.A lot of work was put into ensuring the reliability of NEST. An automated test-suite allows users to check their installation and helps us to detect errors and regressions as new features are added (see companion poster by Eppler et al.) We have also increased the numerical accuracy of many of our neuron models. NEST now supports the recently proposed MUSIC interface [3], to link NEST with different simulators and analysis modules at run-time in a parallel distributed fashion. Since 2007, pre-releases of NEST 2 have been used with great success at international computational neuroscience courses. NEST is released under an open source license for non-commercial use. For more information, please visit our web site at www.nest-initiative.org.NEST is developed by the NEST Initiative, a contract-based collaboration between academic and industrial research institutes. The author presents for the NEST Initiative and would like to thank all its members, particularly (in alphabetic order): Kittel Austvoll, Markus Diesmann, Jochen Martin Eppler, Alexander Hanuschkin, Moritz Helias, Susanne Kunkel, Rüdiger Kupper, Abigail Morrison, Hans Ekkehard Plesser, Tobias Potjans, Wiebke Potjans, and Sven Schrader. Acknowledgements: Research Council of Norway Grant 178892/V3 eNeuro, Next-Generation Supercomputer Project of MEXT, EU Grant 15879 (FACETS), BMBF Grant 01GQ0420, Helmholtz Alliance on Systems Biology