Associative-Projective Neural Networks (APNNs) - Neural Networks and Micromechanics - page 77

Fig. 5.4 Timing diagram of
the associative field

Fig. 5.5 The structure of the
buffer field

neuron activity is transferred from the additional outputs Q *

to the basic outputs

Q and is fixed for the period of the following sync pulse.
Thus, the regulation of the activity of the associative field is accomplished.
A quantity of active neurons of the field always occurs approximately equal to the
established value m and, as a rule, only insignificantly exceeds it.
Buffer neural field .
Let us describe the structure of the buffer field (see Fig. 5.5 ). The neurons of the
buffer field do not use the associative inputs, the threshold set input, or the training
input. One of the set inputs and one of the inhibitory inputs are common for all
neurons of the buffer field (forced installation of neurons into the excited or
inhibited state). Synchronous input is common for all the neurons of the buffer
field. Additional output in the buffer field is not used. The block SUM counts the
number of active neurons in the field. The output of each neuron of the buffer field
is connected with one of the set inputs of the same neuron, making it possible to
accumulate the activity of neurons up to the appearance of a signal of common reset
(which inhibits the activity of the field).