9 oct 1986 año - Paper: Learning representations by back-propagating errors

Descripción:

David E. Rumelhart and Ronald J. Williams of the University of California, San Diego, along with Geoffrey E. Hinton of Carnegie Mellon University, publish a paper in Nature titled “Learning representations by back-propagating errors.” The paper introduces the backpropagation algorithm, a method for training multilayer neural networks by minimizing the difference between actual and desired outputs through iterative weight adjustments. This approach enables hidden layers within the network to develop internal representations that capture important features of the input data, distinguishing it from earlier methods like the perceptron-convergence procedure.

Añadido al timeline:

fecha:

Fotos: