Learning representations by back-propagating errors
The 1986 Nature paper demonstrates backpropagation for learning internal representations in multilayer networks. Its advance over earlier, simpler methods lies in hidden units that create useful new features as weight adjustments reduce the output error.
Overview. The 1986 Nature paper demonstrates backpropagation for learning internal representations in multilayer networks.
Context. Its advance over earlier, simpler methods lies in hidden units that create useful new features as weight adjustments reduce the output error.
Claims and evidence
Claim ledger, 2The 1986 Nature paper demonstrates backpropagation for learning internal representations in multilayer networks.
Its advance over earlier, simpler methods lies in hidden units that create useful new features as weight adjustments reduce the output error.