Model descriptions in Neuroscience: computational performance and collaboration
Résumé
This report reviews the state-of-the-art of model descriptions for largescale neuronal network models in computational neuroscience, with a particular focus on issues of collaborative model development and of performance on large clusters and supercomputers. After summarising the requirements for this class of models, and the capabilities of existing simulation tools and existing model description formats, we analyse the shortcomings of existing tools and languages, and make recommendations for future development in this domain; in particular: the use of mixed, standardised text and binary file formats (e.g. YAML/JSON with parallel HDF5); alternative/interoperable representations to support a wide range of use cases from algorithm-driven point-neuron networks to data-driven biophysically-detailed networks; and the development of conversion tools to allow gradual convergence without disruption to ongoing projects.
Domaines
Complexité [cs.CC]
Fichier principal
Model_descriptions_in_computational_neuroscience__mbox_supercomputers.pdf (1.29 Mo)
Télécharger le fichier
Origine | Fichiers produits par l'(les) auteur(s) |
---|