Information Technology Reference
In-Depth Information
FIGURE 1. Minimal network for computing an elementary conditioned reflex
(ECR). (I) Before conditioning; (II) After conditioning.
Part 1 of Figure 1 shows the minimal net capable of computing an ECR.
Neurons A, B transmit the conditioning and conditioned stimulus respec-
tively to the motoneuron with threshold q=2 -e, where 0 <e<<1, which
always fires when A fires, since the double excitation on its two synapses
override its threshold of less than two units (A single synapse represented
one unit of excitation). The motoneuron cannot be triggered by B only, for
one synapse is insufficient to override its threshold.
However, when first concomitance of A and B occurs, the internuncial is
activated and provides sufficient facilitation for B to initiate the reflex. The
internuncial's recurrent collaterals secure its permanent excitated state and,
henceforth, B only is sufficient to elicit a response.
In spite of the structural simplicity of this four-element network, it
exhibits some features that are instructive in this context. First it should be
noticed that it alters its function as a consequence of the occurrence of
certain stimulus configurations: before concomitance of A with B, the net
is impervious to B, while afterwards it is responsive to B as it is and was to
A. Unfortunately, some authors seem to associate with this simple alter-
ation higher mental functions by calling this “learning through experience”.
Whether this misrepresentation is caused by underestimating the complex-
ity of the processes that establish algorithms for solving certain classes
of problems—i.e., “learning” in its proper sense—or by overestimating the
Search WWH ::
Custom Search