Using neural networks to solve equations?

Question

Programming

So, my goal is to do something similar to WolframAlpha, but only in the mathematical realm. The differences will be that my project is free and will work offline.
I am developing in C++, in Borland C++Builder so far only under Windows. At the moment, I made a calculator that counts step by step, analyzing one line.
Now to the essence of the question:
As far as I can imagine, for solving equations, as in WolframAlpha, there are two ways:
1) Prescribe all options for solving equations manually. I don’t know how realistic this is, because. There is such a variety of equations, and to solve equations of higher degrees, as well as exponential equations, it will be necessary to prescribe very, very complex and large algorithms.
2) Use of Neural networks. I don’t know how to specifically implement them, but I am familiar with the principles of work.
What do you think is the easiest way to solve equations? Do I understand correctly that if you create and configure a neural network, then it will not be necessary to prescribe an algorithm for each type of equation, but will it decide everything on its own? Can neural networks be used offline?
Please, tell us about the optimal, in your opinion, solution to this problem.

Answer 1

The easiest way to solve equations is to learn higher algebra.
It has been proven that the direct logical solution method is always better than a neural network.

Answer 2

Any equations are solved through n-dimensional sets.
The task is to find the extremums of the boundaries of these sets and analyze intersections and touches.
You need to learn how to work with n-dimensional sets.