Presents a strategy for searching patterns 'similar' to those ones presented to and recalled by the trained BP-network but with a lower error at the output of the GREN-network.
