Neural Network Inputs and Outputs to meaningful values

Question

I'm trying to determine how to transform my "meaningful input" into data for an Artificial Neural Network and how to turn the output into "meaningful output".

The way I can always see of doing it is by convering everything to categories with binary values.

For example, rather than outputting age, having a 0-1 for <10, a 0-1 for 10 - 19, etc.

Same with the inputs, where I might be using for example, hair colour. Is the only way to turn this into input to have Blonde 0-1, Brown 0-1, etc?

Am I missing some entire topic of ANNs? Most of the books and similar I read use theoretical examples.

Answer 1

Your example is spot-on, and is known as discretization. Another method would be to scale your continuous input/output into the 0-1 range. For your other example of hair color, that would be a nominal attribute, so the only real option there is to discretize it as separate inputs per color.