Data Science: Backpropagation (Definitions)

"A learning algorithm for multilayer neural nets based on minimizing the mean, or total, squared error." (Laurene V Fausett, "Fundamentals of Neural Networks: Architectures, Algorithms, and Applications", 1994)

"A learning scheme by which a multi-layer feedforward network is organized for pattern recognition or classification utilizing an external teacher, and error feedback (or propagation)." (Guido J Deboeck and Teuvo Kohonen, "Visual explorations in finance with self-organizing maps", 2000)

"weight-vector optimization method used in multilayered feed-forward networks. The corrective steps arc made starting at the output layer and proceeding toward the input layer." (Teuvo Kohonen, "Self-Organizing Maps" 3rd Ed., 2001)

"A class of feed-forward neural networks used for classification, forecasting, and estimation. Backpropagation is the process by which connection weights between neurons are modified using a backward pass of error derivatives." (David Scarborough & Mark J Somers, "Neural Networks in Organizational Research: Applying Pattern Recognition to the Analysis of Organizational Behavior", 2006)

"A method for training a neural network by adjusting the weights using errors between the current prediction and the training set." (Glenn J Myatt, "Making Sense of Data: A Practical Guide to Exploratory Data Analysis and Data Mining", 2006)

"A supervised learning algorithm used to train artificial neural networks, where the network learns from many inputs, similar to the way a child learns to identify a bird from examples of birds and birds attributes." (Eitan Gross, "Stochastic Neural Network Classifiers", 2015)

"A learning algorithm for artificial neural networks used for supervised learning, where connection weights are iteratively updated to decrease the approximation error at the output units." (Ethem Alpaydın, "Machine learning: the new AI", 2016)

"Learning algorithm that optimizes a neural network by gradient descent to minimize a cost function and improve performance." (Terrence J Sejnowski, "The Deep Learning Revolution", 2018)

"The backpropagation algorithm is an ML algorithm used to train neural networks. The algorithm calculates for each neuron in a network the contribution |  the neuron makes to the error of the network. Using this error calculation for each neuron it is possible to update the weights on the inputs to each neuron so as to reduce the overall error of the network. The backpropagation algorithm is so named because it works in a two stage process. In the first stage an instance is input to the network and the information flows forward through the network until the network generates a prediction for that instance. In the second stage the error of the network on that instance is calculated by comparing the network's prediction to the correct output for that instance (as specified by the training data) and then this error is then shared back (or backpropagated) through the neurons in the network on a layer by layer basis beginning at the output layer." (John D Kelleher & Brendan Tierney, "Data Science", 2018)

"Backpropagation is short for 'backward propagation of errors'. Backpropagation in convolutional neural networks is a way of training these networks based on a known, desired output for a specific sample case." (Accenture)