Skip to main content
Providing community access to large-scale neuromorphic computing systems
M. Sénoville, H. Aguili, O. Ates, J. Duperrier, D. Guarino, B. Lungsi Sharma, B. Kindler, E. Müller, A.G.D. Rowley, A.P. Davison, and the EBRAINS Neuromorphic Computing Collaboration
Presenting author:
M. Sénoville
The Neuromorphic Computing Platform of the EBRAINS Research Infrastructure targets researchers in multiple fields, including computational neuroscience and machine learning, and offers them access to two unique large-scale neuromorphic computing systems: the BrainScaleS 20-wafer system in Heidelberg and the SpiNNaker 1,000,000 core machine in Manchester.

With the goal of making the use of these machines as easy as possible, the platform provides multiple access routes:
(i) batch job submission through a graphical interface in the EBRAINS Collaboratory, in Jupyter notebooks, or on the command-line;
(ii) interactive access through EBRAINS-integrated Jupyter notebooks.

Both systems are "programmed" through defining spiking neural networks, with optional synaptic plasticity, using the easy-to-learn, flexible and powerful programming interface, PyNN, which encourages portability and reusability of a model and then facilitates comparisons and crosschecks between neuromorphic computing and traditional software simulators.

Simulation results, with detailed provenance metadata including full log files, are accessible through a graphical interface embedded in the EBRAINS Collaboratory.

This poster will present the current status of the platform, with a focus on recent developments and upcoming features.