How is backpropagation different from stochastic gradient descent?

Question

Neural networks

As far as I understood from the literature, they differ only in that the backpropagation of the error changes after processing the entire sample, and the stochastic descent after each image?

Answer 1

In fact, they are slightly different concepts.
Backpropagation gives us a function to optimize ( = find min\max).
SGD finds the minimum/maximum of some arbitrary function.
In addition to stochastic descent, there is also full-batch gradient descent (this is the one, by default, when we change the coefficients after calculating the entire sample), mini-batch gradient descent (an intermediate option, when we change the coefficients after calculating N sample elements), rmsprop, ADAM , simulated annealing, etc.