Arguably the most significant contribution of the pseudo-recurrent memory model is its
ability to learn data in a serial manner without experiencing severe interference
(Hetherington, 1991). In this experiment, the patterns were taken randomly from the 1984
Congressional Voting Records database at the UCI repository. Twenty members of
Congress (10 Republicans, 10 Democrats, defined by their voting record on 16 issues and
their party affiliation) were chosen randomly from the database and were presented
sequentially to the network. The order of presentation was randomized in each run of the
program and the results averaged over 50 runs