Abstract
Since the beginning of information processing by electronic components, thenervous system has served as a metaphor for the organization of computationalprimitives. Brain-inspired computing today encompasses a class of approachesranging from using novel nano-devices for computation to research intolarge-scale neuromorphic architectures, such as TrueNorth, SpiNNaker,BrainScaleS, Tianjic, and Loihi. While implementation details differ, spikingneural networks -- sometimes referred to as the third generation of neuralnetworks -- are the common abstraction used to model computation with suchsystems. Here we describe the second generation of the BrainScaleS neuromorphicarchitecture, emphasizing applications enabled by this architecture. Itcombines a custom analog accelerator core supporting the accelerated physicalemulation of bio-inspired spiking neural network primitives with a tightlycoupled digital processor and a digital event-routing network.