A foundational work in the field of neural networks. It introduces the concept of backpropagation, a method for efficiently training multi-layer neural networks through the adjustment of weights based on the gradient of the error function. This technique allows networks to learn complex representations of data by minimizing errors in a systematic way. Backpropagation has become a central algorithm in deep learning, enabling significant advancements in areas such as image recognition, natural language processing, and beyond.
Learning representations by back-propagating errors
Authors: David E. Rumelhart, Geoffrey E. Hinton, Ronald J. Williams
Date: