Is it more accurate to solve a problem with the help of AI?

Question

Mathematics

In general, it is very necessary to teach the ANN to solve, let's start with a simple, ordinary equation with one unknown, that is, of the form a + x \u003d b, where a, b are the numbers that will be given. How many neurons are needed for this, layers of the learning algorithm, training sample, testing sample. In the future, I plan to teach how to solve a quadratic equation, then a differential one, and finally the Schrödinger equation. Explain to an inexperienced professional

Answer 1

A neural network is often used as a function (graphics) approximator. To begin with, I would rewrite your equation (change of variables):
x1 + y = x2
We get the function
y = -x1 + x2.
This is a neural network with inputs x1 and x2 and output y.
One neuron, the weight of the first input is -1, the weight of the second input is 1.
The activation function is linear.
The answer is ready, no need to teach.