Common control principles of basal ganglia - thalamocortical loops and the hippocampus
A. Lõrincz
Neural Network World,
in press (1997)
Abstract
The learning features of the hippocampal formation and
the basal ganglia - thalamocortical
loops are analysed on the basis of their
recent models utilizing control concepts for both
architectures.
According to the model of the trisynaptic loop of the hippocampus
independent components of long term memory are trained by this
loop and during the consolidation of these memories they are temporally
locked by the same loop. Temporal locking, or temporal compensation,
is achieved by control means.
It is further argued that the control signal develops
statistically independent components,
important for goal oriented systems since independent components
promote generalization and higher order concept formation.
The control model of the basal ganglia - thalamocortical loops is
based on a similar architecture. It exhibits
global stability and on-line tunability if certain structural conditions
are met. It is argued that skill learning forms a hierarchy that
first identifies `sign-proper' feedback channels and then tunes these
channels. These channels can control and can also serve
as error feedback correcting means for
an external motor programme generator. The possibility of
skill learning for hippocampal subjects is discussed.
