From the course: DJ Patil: Ask Me Anything

What are neural networks (NN)?

From the course: DJ Patil: Ask Me Anything

Start my 1-month free trial

What are neural networks (NN)?

“

(upbeat music) - [DJ] What are neural networks? The easiest way to think about a neural network is think about a bunch of nodes. Think about a bunch of nodes, you've got node one, node two, node three, and nodes, and they are all connected somehow. And those nodes start at a certain base state and the connection between these nodes, think of 'em like the line between them is like a highway or a road, and some of those roads are a highway really built out. There's lots of traffic going through it, or there's a road that's just barely, somebody walks, occasionally it's overgrown. And then what you do is you show this graph, this network, and you say basically, "Hey, this is an image," or "This is something that I want you to test it on." And based on that, these nodes represent certain factors. If this was a neural network for an image, we might have a node that says "This is light or dark," another node that might have an idea of the contrast, or this blue, red. And so you can have all of these nodes. Now, if we showed a certain amount, nodes that are connected together might get more connection, so they're thicker and over time, there's some pattern that evolves through this. And so now this network, when it sees something that says, "Ah-ha, it's this type," it effectively recognizes it, and so it pattern-matches. It says, 'Oh yes, that's a bright image with the sun in it." Or, "Oh, that's a cloudy image that has this," because of the way the network actually is tied together. That's the, sort of, the simplest way to think of it. (upbeat music)

Contents