Neural Nets And Why They Matter
What is a Neural Net and why does it matter to a normal person ?
There are plenty of online resources to search on Neural Nets, so instead I'll expand on why it matters to a normal person.
The words or concepts that will be explained a little bit here are
- network
- train or trained
- sensory inputs as relates to a neural net
How many things can either of us do at any one time ? Probably just a few, at least do well. And life is all about doing, when it comes down to results.
So, What Gets Our Attention ?
If you think of attention as an on/off switch, Neural Nets make everything logical.
First, lets forget that we are humans to simplify the picture, and just note that we are also mammals. What can a mammal sense ?
A mammal has 5 channels to sense ( seeing, hearing, smelling, touch, and taste). Some mammals are better in certain categories, for example a dog can smell a million times better than us ?? An eagle can see more movement at long distances but less clear detail ??
All senses are on, all the time. So there has to be some kind of filter mechanism for getting attention, for figuring out which things matter, which need to be acted on for survival, etc. That's pretty easy to understand. Even an 8 year old has learned that he has to pay attention to the teacher.
What is A Trained Network in this Context ?
A Neural Net is a way of understanding the world in terms of trained networks. In this case, a network is a fancy way of saying "Dude, you've got all these senses turned on all time, so there are lots of inputs coming in at all times on all senses, so the sum total of all these inputs coming in has to have a name. We'll call that a 'network'."
Since there too many things going on, let's go back to the 8 year old sitting in class paying attention to the guys playing soccer in the playground out the window. His network is being trained. That teacher better do a more interesting job, otherwise the poor guy is really going to mis-train his network and pay attention to the guys in the playground playing soccer, and his network may (over time) grow to be mis-aligned.
When the network is training or trained, it becomes increasingly comfortable with certain types of information. One of the most common examples used in Neural Network academics circles is the face. Mammals such as humans acquire the ability to recognize faces, that is a "trained" neural net capability.
Why Does It Matter ?
There are two reasons why it matters to the average person.
- You're not going to "see" important stuff if your Neural Net isn't "trained".
- You're going to gravitate towards weird stuff if you don't understand training.
What is important stuff and why would you miss it ? What is weird stuff and why would you gravitate towards it ?
About Missing Important Stuff
It was told that the so called "Indians" that inhabited the American continents did not even see the ships of the invading Europeans as they were sailing in. They didn't know about such ships, so even seeing them on the distant horizon, they wouldn't pick them up visually. Their neural nets were not trained.
Another example: Scientists often miss important discoveries because the resulting experimental data just doesn't fit their pre-conceived notions. So they dismiss it, and some other scientist "discovers" what they had already seen in their test data, and just hadn't filtered out.
How many things fit this pattern in our lives ? For me, being a 55 year old I could recite an embarrassing list of things that were quite obvious that I missed entirely for decades. How about you ?
About Gravitating Towards Weird Stuff ?
OK, maybe weird is just a word I used to call attention to the nature of certain senses. Because it really isn't so weird at all.
One friend posts amazement at how much art is used to call attention to sadness and tragedy. But is that really so odd ? Music that is sad is going to appeal to someone who has a neural net trained towards sadness by all the other events in his life. All the more so, if the person is a bit unsettled, and the music pulls it all together for him.
Consider a much more tragic example. A woman who is accustomed to being abused and leaves an abusive relationship will more often than not simply enter another abusive relationship. This is not just about abused women though, we all gravitate towards the familiar, no matter how tragic that familiar is.
Think of it as a Flood
Think of a neural net as a flood of information. With all 5 senses on at all times, your system has to be able to focus on more like one thing at a time. So it uses the same pipelines of sensory inputs that it used yesterday and 10 minutes ago, and it discriminates by which ones get flooded most often and how. We learn to pick up floods of information in certain areas, by training.
If the flow of information is big enough to flood the pipes and get a familiar read, then it turns a switch of our attention on. If it doesn't, we're going to pay attention to something else. That is either good or bad or it doesn't matter, but either way, the flood is what controls the attention circuit.
That may be a gross oversimplification of a Neural Net. I hope that it helps you conceptualize a bit better, because the whole thing about Neural Nets confused the heck out of me for a long time. But it really isn't that complicated.
