Neural nets

I’ve always had a problem with neural nets. They claim to be a learning system, but I’ve never got ‘the learning’ part of it.

Well…I think I came up with the magic piece that I was missing.

I couldn’t understand how learning could be in stored in a pathway…each node with a distinct piece of information…I mean how do you make a pathway say ‘Dog’?

Now, my current thinking is that the pathways aren’t distinct. They’re weighted. The strength of a node’s charge is based on the ‘weight’ or current that it is carrying. The more supporting nodes you have, the stronger you are. The stronger you are, the more likely that your pathway will be chosen.

Each neural node gets broken up into sub-nodes, each with additional information about the main node. For example, I can press the middle C on a piano, but that is a world apart from being a concert pianist touching a the same middle C. The pianist has practiced to the point that there are hundreds of sub-nodes under that middle C, refining the fingers to the point that they play the key with just the right pressure and just the right time.