A Learning Algorithm for Continually Running Fully Recurrent Neural Networks

Neural Computation - Tập 1 Số 2 - Trang 270-280 - 1989
Ronald J. Williams1, David Zipser2
1College of Computer Science, Northeastern University, Boston, MA 02115, USA
2Institute for Cognitive Science, University of California, La Jolla, CA 92093, USA#TAB#

Tóm tắt

The exact form of a gradient-following learning algorithm for completely recurrent networks running in continually sampled time is derived and used as the basis for practical algorithms for temporal supervised learning tasks. These algorithms have (1) the advantage that they do not require a precisely defined training interval, operating while the network runs; and (2) the disadvantage that they require nonlocal communication in the network being trained and are computationally expensive. These algorithms allow networks having recurrent connections to learn complex tasks that require the retention of information over time periods having either fixed or indefinite length.

Từ khóa


Tài liệu tham khảo

Almeida L.B., 1987, Proceedings of the I E E E First International Conference on Neural Networks, 11, 609

10.1073/pnas.79.8.2554

Jordan M.I., 1986, Proceedings of the Eighth Annual Conference of the Cognitive Science Society, 531

10.1016/0167-2789(86)90244-7

10.1109/TAC.1965.1098164

10.1016/0885-064X(88)90021-0

Rohwer R., 1987, Proceedings of the I E E E First International Conference on Neural Networks, 11, 701

Stornetta W.S., 1987, Proceedings of the I E E E Conference on Neural Information Processing Systems, 750

Thông tin
Thông tin xuất bản

Neural Computation

Tập 1 Số 2

270-280

Thông tin tác giả