Logo of the Human Brain Project

BrainScaleS-1 demonstrates wafer-scale neuromorphic computing

As part of his thesis Implementation of Large Scale Neural Networks on the Neuromorphic BrainScaleS-1 System Jakob Kaiser successfully implemented the largest neural network to date with 230 — of 383 available — HICANN chips operating simultaneously making full use of wafer-scale integration. The feed-forward chain studied, see raster plot, includes feed forward inhibition and consists of 190 chain links of 100 logical neurons each. 76000 neuron circuits and 1.4 million synapses were involved in the experiment. The neural dynamics evolve with a speed-up factor of 10000 compared to the biological time-scales the systems emulates.

Last update of this page: 2020-09-30 by Jakob Kaiser