- Latent Connection is a Netlab term with no direct analog in the study of biological neural networks.
Latent Connections or LC's are possible connections defined in the synapse
space of a neuron
between itself and other specific neurons. That is, a Latent Connection
is a predefined possible synapse connection between two specific neurons (or between any two devices in Netlab).
Difference between LC's and RPs
As stated, an LC is a predefined (possible) connection between a specified
target neuron and the axon
(output) of another specified
neuron. The key word is 'specified'
. The specific pre- and post-synaptic neurons being connected are determined at design-time. The latent connection can develop into a full connection based on how the weights
are adjusted at run time, but it is limited to being between the two neurons (pre and post), which had been specified
is a specified site on a target neuron that can form a dynamic connection to a single axon
from any source within the system (depending on run-time experiences). The key word here is 'any'
. The pre-synaptic
neuron that makes connection to an RP
is determined at run time and is not specified at design time.
s are for specifying a possible connection between a specified axon and a specified neuron (or bundle). These are what is thought of when considering typical ANN
topologies. They are useful in many networks that only need hard coded, latent organization.
s are places on a neuron that may form a connection with any other neuron in the system at run-time. In Netlab
s may be used exclusively to build networks, or may be designed as part of the structure of networks that also include LC
s are part of Netlab's pathfinding
functionality, which includes pathfinders and chemical influences