The learning can be considered to have succeeded if it its output pattern is the same as the input pattern.
This study examined how much different a pattern can be fed into the network that completed the learning.
The network which consist of 100, 200, 300, 400 elements ware prepared, and the random pattern of 1 and -1 binary was learned in the network.
It was examined whether the network which completed the learning could succeed a recall by feeding a little different experimental input pattern.
The experimental input pattern was generated by reversing the bits of a learned pattern, and the number of bits were increased.
As a result, it has been understood that the number of bits in which the recalls succeed decreases when the learned pattern increases.
At most 50% of reversed bit was possible to recall in any element numbers, as long as the learned pattern is one.
And, at most 2% of reversed bit was possible to recall in any element numbers,when the learned pattern is the maximum complete learned pattern.