Some GREN-networks lack in their ability to train BP-networks. In this paper, we propose a strategy for increasing their sensitivity to BP-network outputs.
