are terms used to describe the connection semantics of a synapse
. A synapse is a connection between two entities, one sourcing the connection with its axon
: the pre-synaptic
entity, and one receiving the axon connection: The post-synaptic entity
While shown for a single connection this works just as well when describing or discussing multiple entities that are party to any number of connections.
This terminology is preferred over the ANN
convention of speaking in terms of hidden
neurons for at least two reasons.
- It prevents inadvertently pulling the entire system into the discussion, and
- it doesn't box in our discussion around feed-forward-only restrictions that were once imposed by earlier ANN learning algorithms. For an example of a learning algorithm that doesn't impose any feed-forward restriction, or any restrictions on feedback for that matter, see Influence Learning