[Rick]

For those of you that don't understand:

You have a bunch of data and you feed it into a computer, in the example used in OP this is height and weight.

Each time you feed it in you tell the computer "This is a dog" or "This is a cat" or "This is a bird".

Eventually using analysis systems built into the program it can build up some paramaters that allow it to recognise whether the data you give it is a cat dog or bird without you telling it.



It's just like a very core concept of learning. When you are young you use this exact same method but you learn quicker because you have so many more paramaters given to you other than Weight and Height.

If I read this thread correctly, this is relevant:

Neuro-Evolution or whatever he calls it in this is pretty neat. It's good because the computer will be told "That worked well" or "That worked bad" and it changes its behavior based on feedback that you don't directly give it, it can learn to be perfect by itself given enough time.

It's a lot more efficient than the taught system in the OP. The OP example is like teaching a child. Whereas the Neuro-Evolution is like a child learning a new skill by itself.