If you would like to know how exactly it works here is part one

Right now as I'm writing this I'm on the bus so mind my spelling and Grammer.

First what happens is you look at the line you want to make. You will have a library full of equations and patterns and you will choose what patients or equation seems to fit well or has keywords that matches the line you want to match. After that you will get the equations and modify there variables to match sections in the line you are trying to match. You repeat this step modifying and translating the graph.

After this you can take the error the line you made in relation to the line you want to match. Now find the line of best for the error. Then you add those two together to get the new equation.

Some questions you may have are:

1. How is this more effective at training. Answer: Well with this instead of having to basically guess how many neurons you would need in your nn; you will just be able to map it. This will also help against over fitting or overcomplicating a network. It can also help you run bigger networks on less powerful devices.
2. How will I know what equation to use? Answer: in the library it will have multiple sections including images each aqation can give you and more on how they can modify. This way you can easily find equations you need.
3. Wouldn't it be better to just use simpler equations? Answer: in allot of cases no, each neuron in this method will have more control individually and can most cases simplify networks and make it easier for non so devs to learn.
4. How would this work on 2 inputs network or more? Answer: this is still in development but will be explained in my next post.