How does a neural network form hidden layers?

Question

Neural networks

Here's what I understand: There is a neural network with 2 hidden layers, for example, for face recognition. A photo of a face is given as input, each pixel is a separate feature, all features make up the first layer of the neural network. Then there is the second layer on which the input features are *combined* into basic primitives using non-linear functions. Further, on the 3rd layer, the primitives are combined into more abstract things, and the output is the answer to the face in the picture or not.
Question: How does the neural network determine those very primitives in the hidden layers? How does she understand that these two lines make it clear if there is a nose in the picture, why did she choose the nose as an abstraction in principle?
Thanks

Answer 1

Have you tried to find the answer yourself? After all, there is more than enough information on the web. Why ask on the forum the question that is comprehensively and in detail considered in books and even just on the Internet? All the same, it is better here than in the prepared source they will not explain.
Well, if you want the necessary sources to be found and presented for you - well, here they are, a small part. Learn:
https://habr.com/en/post/461365/
https://towardsdatascience.com/object-detection-wi...
https://www.kdnuggets.com/2019/08/2019-guide- objec...
https://www.pyimagesearch.com/2018/06/18/face-reco...
https://towardsdatascience.com/introduction-to-ima...
https://towardsdatascience.com/ how-to-detect-objec...
https://towardsdatascience.com/a-beginners-guide-t...
https://towardsdatascience.com/computer-vision-cre...
https://towardsdatascience.com/face-recognition-us...