Extracting finite-state representations from recurrent neural networks trained on chaotic symbolic sequences | IEEE Journals & Magazine | IEEE Xplore