PyNN for BrainScaleS-2

BrainScaleS-2 allows users to use the PyNN API to define experiments of spiking neural network. This documentation provides details to the BrainScaleS-2 implementation of PyNN and highlights differences to the standard PyNN interface. More details to the PyNN API can be found in the corresponding documentation.

BrainScaleS-2 is an accelerated, mixed-signal neuromorphic chip; its analog circuits implement the dynamics of the adaptive exponential integrate-and-fire neuron model. The custom cell type HXNeuron allows to set the “hardware parameters” of neuron circuits directly. For more information about the HXNeuron see the corresponding documentation.

In addition the cell types SpikeSourceArray and SpikeSourcePoisson are available to inject external spikes into the network. Just like in standard PyNN, populations of neurons and projections between populations are used to define the network architecture. A good starting point to get familiar with BrainScales2 and its PyNN interface are the demos and tutorials.

Before the network is emulated on the BrainSacaleS-2 system, the abstract network description has to be translated to a valid hardware configuration. This mapping is performed by grenade

Recording and receiving of observables such as spikes and membrane voltages work as in standard PyNN. For a list of recordable observables refer to the HX neuron documentation.

API Overview

An overview over the full API can be found in /api_pynn-brainscales2.