In an attempt to incorporate the two new patterns into network without disrupting the 12
previously learned patterns, various learning rates, ranging from η=0.01 to η=0.0001 with
various values for the momentum term, ranging from 0.05 to 0.9, were tested. Once the
network had learned the two new (disruptive) patterns (often quite slowly because of the low
learning rate), it was then tested on how well it remembered the previously learned set of
twelve associations. Retraining on the original data was again done with a learning rate of 0.2
with momentum 0.9. Both the Ebbinghaus-savings and the exact-recognition of forgetting were used