What does "backprop" mean? Is the "backprop" term basically the same as "backpropagation" or does it have a different meaning?


5 Answers 5


"Backprop" is the same as "backpropagation": it's just a shorter way to say it. It is sometimes abbreviated as "BP".


'Backprop' is short for 'backpropagation of error' in order to avoid confusion when using backpropagation term.

Basically backpropagation refers to the method for computing the gradient of the case-wise error function with respect to the weights for a feedforward networkWerbos. And backprop refers to a training method that uses backpropagation to compute the gradient.

So we can say that a backprop network is a feedforward network trained by backpropagation.

The 'standard backprop' term is a euphemism for the generalized delta rule which is most widely used supervised training method.

Source: What is backprop? at FAQ of Usenet newsgroup comp.ai.neural-nets


  • Werbos, P. J. (1974). Beyond Regression: New Tools for Prediction and Analysis in the Behavioral Sciences. PhD thesis, Harvard University.
  • Werbos, P. J. (1994). The Roots of Backpropagation: From Ordered Derivatives to Neural Networks and Political Forecasting,Wiley Interscience.
  • Bertsekas, D. P. (1995), Nonlinear Programming, Belmont, MA: Athena Scientific, ISBN 1-886529-14-0.
  • Bertsekas, D. P. and Tsitsiklis, J. N. (1996), Neuro-Dynamic Programming, Belmont, MA: Athena Scientific, ISBN 1-886529-10-8.
  • Polyak, B.T. (1964), "Some methods of speeding up the convergence of iteration methods," Z. Vycisl. Mat. i Mat. Fiz., 4, 1-17.
  • Polyak, B.T. (1987), Introduction to Optimization, NY: Optimization Software, Inc.
  • Reed, R.D., and Marks, R.J, II (1999), Neural Smithing: Supervised Learning in Feedforward Artificial Neural Networks, Cambridge, MA: The MIT Press, ISBN 0-262-18190-8.
  • Rumelhart, D.E., Hinton, G.E., and Williams, R.J. (1986), "Learning internal representations by error propagation", in Rumelhart, D.E. and McClelland, J. L., eds. (1986), Parallel Distributed Processing: Explorations in the Microstructure of Cognition, Volume 1, 318-362, Cambridge, MA: The MIT Press.
  • Werbos, P.J. (1974/1994), The Roots of Backpropagation, NY: John Wiley & Sons. Includes Werbos's 1974 Harvard Ph.D. thesis, Beyond Regression.
  • $\begingroup$ maybe you need to accept your own answer. $\endgroup$ Commented Sep 27, 2021 at 1:29

Yes, as Franck has rightly put, "backprop" means backpropogation, which is frequently used in the domain of neural networks for error optimization.

For a detailed explanation, I would point out this tutorial on the concept of backpropogation by a very good book of Michael Nielsen.


It's a fancy name for the multivariable chain rule.

  • 1
    $\begingroup$ Hello. Could you support this answer with a little more detail? (I realize it's a simple question, but I think your answer is worthy of greater explication.) $\endgroup$
    – DukeZhou
    Commented Apr 23, 2020 at 21:06
  • $\begingroup$ @FourierFlux Please, follow the suggestion in the previous comment. At least, provide a link to an external resource, such as Wiki, that provides more details. Take a look at ai.stackexchange.com/help/how-to-answer to know more about what we expect from answers and how to write a good answer. $\endgroup$
    – nbro
    Commented Jan 10, 2021 at 0:30

We need to compute the gradients in-order to train the deep neural networks. Deep neural network consists of many layers. Weight parameters are present between the layers. Since we need to compute the gradients of loss function for each weight, we use an algorithm called backprop. It is an abbreviation for backpropagation, which is also called as error backpropagation or reverse differentiation.

It can be understood well from the following paragraph taken from Neural Networks and Neural Language Models

For deep networks, computing the gradients for each weight is much more complex,since we are computing the derivative with respect to weight parameters that appear all the way back in the very early layers of the network, even though the loss is computed only at the very end of the network.The solution to computing this gradient is an algorithm called error backpropagation or backprop. While backprop was invented for neural networks, it turns out to be the same as a more general procedure called backward differentiation, which depends on the notion of computation graphs.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .