David E. Rumelhart, Geoffrey E. Hinton, & Ronald J. Williams (1986)
Nature, 323(6088), 533-536.
DOI: https://doi.org/10.1038/323533a0
Abstract. The Nature paper that popularised backpropagation as a practical training algorithm for multilayer neural networks. Its demonstration that gradient descent via the chain rule could train hidden representations sparked the connectionist revival of the late 1980s.
Tags: neural-networks backpropagation history