How is a neural network trained?

Question

Neural networks

I am writing my "toy" neural network.
But I can’t understand how it is trained, how would I understand that it is back propagation of an error.
And I made this spread.
But I can't understand how to do something like this:
input: [0, 1], output: [0]
input: [1, 0], output: [1]
:'(
the output always turns out to be 1.
Please help.
Thank you.

Answer 1

Your question is taken out of context and without a fundamental formulation of the problem it turns out to be general.
It is important to understand that even a properly built and trained network will give the probability: input: [0, 1], output: [0.11] and input: [1, 0], output: [0.99], and after applying the confidence threshold coefficient to the network response we get the expected result, for example:

parsePrediction(prediction, threshold = 0.5)  {
  if (prediction > threshold) {
    return 1
  } else {
    return 0
  }
}

Look at this implementation, of which there are really many, try to fly an alternative implementation and understanding will come.