560 prohibited the use of probabilistic connection rules when connecting to devices to address #351.
This rule is overly restrictive: Only connection rules that have a global constraint, such as fixed in- or outdegree and fixed total number. In contrast, pairwise Bernoulli checks each connection independently, so as long as each VP only considers sources on that VP, it is safe to use it for connecting to devices.
A use case is to randomly connect a given percentage of neurons in a population to a spike recorder for recording, when recording of all neurons is not practical. This would be especially useful for NEST Desktop. Note that slicing a subset of neurons is only sensible for random networks (mentioned by @babsey).
560 prohibited the use of probabilistic connection rules when connecting to devices to address #351.
This rule is overly restrictive: Only connection rules that have a global constraint, such as fixed in- or outdegree and fixed total number. In contrast, pairwise Bernoulli checks each connection independently, so as long as each VP only considers sources on that VP, it is safe to use it for connecting to devices.
A use case is to randomly connect a given percentage of neurons in a population to a spike recorder for recording, when recording of all neurons is not practical. This would be especially useful for NEST Desktop. Note that slicing a subset of neurons is only sensible for random networks (mentioned by @babsey).