How to create a neural network structure on pluses for handwritten digit recognition?

Question

Neural networks

I can not understand where the weights come from? And how are they distributed throughout the network, in each layer for each neuron? And what should it look like on the layer? It was not difficult to write a neuron and layers with neurons. But it depended on the transmission of signals and communication with neurons among themselves. Plus, I can't figure it out with weights. Chew please. Watched a lot of videos, googled a lot!

Answer 1

if a cactus is tasty for you, it does not mean that it is tasty for others
here and on sugar (Python) there is where to move and what to spend time on ...
here it is chewed

Answer 2

I can not understand where the weights come from? And how they are distributed over the network,

Actually, this is the basis of what is called a "neural network". The search for weights and their "distribution over the network" is, in fact, an optimization problem solved by one of the numerical methods, usually gradient optimization.
If you ask such a question, it means that you are still completely "out of touch". Start with the alphabet, then - when you understand how it all works - you can think about how it is used in "handwritten digit recognition", and only then - how to implement it in C ++. Although why? It is on the pluses that the same Tensorflow and PyTorch have already been implemented. Do you want to compete with them?