Permutation Coding Technique for Image Recognition System - Neural Networks and Micromechanics - page 51

T 21

T 11

RE
R
DE
D
SE
S
Clk

a 1 RE

R
DE
D
SE
S
Clk

Q 11 Q 21

OR

T 22

T 12

a 2

RE
R
DE
D
SE
S
Clk

RE
R
DE
D
SE
S
Clk

Q 12 Q 22

OR

T 23

T 13

a 3

RE
R
DE
D
SE
S
Clk

RE
R
DE
D
SE
S
Clk

Q 23

Q 13

OR

Fig. 4.3 Structure of three associative elements

All the associative elements have the same internal structure, so we describe just
the first associative element as an example. The associative element contains at the
input an OR neuron connected randomly to several neurons of the D 2 layers.
The output of the OR neuron is connected to the S -input of the D -trigger T 11 and
to the D -input of the trigger T 21 . The output of trigger T 11 is connected to the
D -input randomly selected in the same layer of triggers (with T 12 in Fig. 4.3 ). The
output of the trigger T 21 serves as the output of the associative element.
The associative element works as follows. At the beginning, a reset pulse is
applied to the trigger T 11 . After that, a clock pulse is applied to the triggers T 11 and
T 21 . After these pulses, the outputs Q 11 and Q 21 are equal to the output of the OR
neuron. After this procedure, the process of normalization begins, which consists of
several cycles. Each cycle starts from the pulses applied to the inputs Data Enable
(DE) and Clock (Clk) of the trigger T 11 . These pulses lead to permutations of the
outputs of the triggers T 1i . Reset Enable (RE) and Clock (Clk) of trigger T 21 leads to