This is the ensemble connection where the neurons of one ensemble are connected directly to the neurons of another ensemble (very different from a direct connection, where ensemble do not use neurons). These connections do not use encoders or decoders.
The biggest issue with this type of connection is the fact that the data passed around is huge. Thus, it could potentially be a problem to pass this data around using zmq sockets.
If the amount of data being communicated does not affect the node's ability to communicate and does not affect performance, then the ensembles participating in this connection should remain as separate processes. Otherwise, as Terry has suggested earlier, we should place them on a single process.
This is the ensemble connection where the neurons of one ensemble are connected directly to the neurons of another ensemble (very different from a direct connection, where ensemble do not use neurons). These connections do not use encoders or decoders.
The biggest issue with this type of connection is the fact that the data passed around is huge. Thus, it could potentially be a problem to pass this data around using zmq sockets. If the amount of data being communicated does not affect the node's ability to communicate and does not affect performance, then the ensembles participating in this connection should remain as separate processes. Otherwise, as Terry has suggested earlier, we should place them on a single process.