Hill climbing in recurrent neural networks for learning the a/sup n/b/sup n/c/sup n/ language | IEEE Conference Publication | IEEE Xplore