This title appears in the Scientific Report :
2016
NEST-MC: A morphologically detailed neural network simulator for many core high performance computer architectures
NEST-MC: A morphologically detailed neural network simulator for many core high performance computer architectures
Keywords: nest; multicompartment; neurosimulator; many-core; HPC; HBP. The nest-mc multicompartment neural network simulator will enable new scales and classes of mor- phologically detailed network simulations on current and future supercomputing architectures. Nest-mc is being developed as a collab...
Saved in:
Personal Name(s): | Klijn, Wouter (Corresponding author) |
---|---|
Cumming, Benjamin / Karakas, Vasileos / Peyser, Alexander / Yates, Stuart | |
Contributing Institute: |
Jülich Supercomputing Center; JSC |
Imprint: |
2016
|
Conference: | IAS Symposium 2016, Jülich (Germany), 2016-12-05 - 2016-12-06 |
Document Type: |
Poster |
Research Program: |
Supercomputing and Modelling for the Human Brain Theory, modelling and simulation Computational Science and Mathematical Methods |
Publikationsportal JuSER |
Keywords: nest; multicompartment; neurosimulator; many-core; HPC; HBP. The nest-mc multicompartment neural network simulator will enable new scales and classes of mor- phologically detailed network simulations on current and future supercomputing architectures. Nest-mc is being developed as a collaboration between the Neuroscience SimLab at the Forschungszentrum Juelich, Barcelona Supercomputing Center and the Swiss National Supercomputing Center (CSCS) under the aegis of the NEST Initiative. The trend towards "many-core" architectures such as GPU and Intel Xeon Phi based systems demands new approaches in software development and algorithm design. Nest-mc is be- ing written specifically for these architectures; it aims to be a flexible platform for neural network simulation, interoperable with models and workflows of NEST and NEURON. Improvements in performance and flexibility will enable a variety of novel experiments, but the design isn’t finalised, and will be driven by the requirements of the community. This is where you come in! We are very interested in your ideas for features which will make new science possible: we ask you to think outside of the box and build this next generation neurosimulator together with us. Possible features and use cases include: * Simulating significantly larger networks over longer time scales: simulate a larger proportion of CNS systems with morphological detail, run longer simulations for slowly developing phenomenon and improve statistical power by leveraging large data sets. * A well-defined high performance C++ API which allows tight integration with other codes: simulate at multiple scales by coupling with other simulators, perform real-time visualization on HPC resources, run online statistics to avoid scaling bottlenecks and embed networks in physically modeled animals. * Dynamic data structures which allow the creation of models with a time-varying number of neurons, synapses and compartments: simulate neuronal development, healing after injury and age related neuronal degeneration. What questions haven’t you asked yet? |