Experiment 2 uses a real-world database to establish that the pseudo-recurrent network
succeeds in significantly reducing interference. Since, in the catastrophic forgetting literature,
two measures of forgetting are frequently used, an Ebbinghaus-like “savings” measure and a
measure of percentage of explicit recognition of previously learned items, we demonstrate in
this experiment that for both of these measures, interference from newly learned patterns is
significantly reduced in the pseudo-recurrent memory model