0:01

The term, Deep Learning, refers to training Neural Networks,

Â sometimes very large Neural Networks.

Â So what exactly is a Neural Network?

Â In this video, let's try to give you some of the basic intuitions.

Â 0:12

Let's start to the Housing Price Prediction example.

Â Let's say you have a data sets with six houses, so you know the size of the houses

Â in square feet or square meters and you know the price of the house and you want

Â to fit a function to predict the price of the houses, the function of the size.

Â So if you are familiar with linear regression you might say, well let's

Â put a straight line to these data so and we get a straight line like that.

Â But to be Pathans you might say well we know that prices

Â can never be negative, right.

Â So instead of the straight line fit which eventually will become negative,

Â let's bend the curve here.

Â So it just ends up zero here.

Â So this thick blue line ends up being your function for

Â predicting the price of the house as a function of this size.

Â Whereas zero here and then there's a straight line fit to the right.

Â 1:17

We have as the input to the neural network the size of a house which one we call x.

Â It goes into this node, this little circle and

Â then it outputs the price which we call y.

Â So this little circle, which is a single neuron in a neural network,

Â implements this function that we drew on the left.

Â 1:53

And by the way in the neural network literature, you see this function a lot.

Â This function which goes to zero sometimes and

Â then it'll takes of as a straight line.

Â This function is called a ReLU function which stands for

Â rectified linear units.

Â So R-E-L-U. And

Â rectify just means taking a max of 0 which is why you get a function shape like this.

Â 2:23

You don't need to worry about ReLU units for

Â now but it's just something you see again later in this course.

Â So if this is a single neuron, neural network,

Â really a tiny little neural network, a larger neural network

Â is then formed by taking many of the single neurons and stacking them together.

Â So, if you think of this neuron that's being like a single Lego brick, you then

Â get a bigger neural network by stacking together many of these Lego bricks.

Â Let's see an example.

Â 2:57

Letâ€™s say that instead of predicting the price of a house just from the size,

Â you now have other features.

Â You know other things about the host, such as the number of bedrooms,

Â I should have wrote [INAUDIBLE] bedrooms, and you might think that one of the things

Â that really affects the price of a house is family size, right?

Â So can this house fit your family of three, or family of four, or

Â family of five?

Â And it's really based on the size in square feet or square meters, and

Â the number of bedrooms that determines whether or

Â not a house can fit your family's family size.

Â And then maybe you know the zip codes,

Â in different countries it's called a postal code of a house.

Â And the zip code maybe as a future to tells you, walkability?

Â So is this neighborhood highly walkable?

Â Thing just walks the grocery store?

Â Walk the school?

Â Do you need to drive?

Â And some people prefer highly walkable neighborhoods.

Â And then the zip code as well as the wealth maybe tells you, right.

Â Certainly in the United States but some other countries as well.

Â Tells you how good is the school quality.

Â So each of these little circles I'm drawing, can be one of those ReLU,

Â rectified linear units or some other slightly non linear function.

Â So that based on the size and number of bedrooms,

Â you can estimate the family size, their zip code, based on walkability,

Â based on zip code and wealth can estimate the school quality.

Â And then finally you might think that well the way people decide how much they're

Â will to pay for a house, is they look at the things that really matter to them.

Â In this case family size, walkability, and school quality and

Â that helps you predict the price.

Â 4:57

And so by stacking together a few of the single neurons or the simple predictors

Â we have from the previous slide, we now have a slightly larger neural network.

Â How you manage neural network is that when you implement it,

Â you need to give it just the input x and

Â the output y for a number of examples in your training set and

Â all this things in the middle, they will figure out by itself.

Â 5:25

So what you actually implement is this.

Â Where, here, you have a neural network with four inputs.

Â So the input features might be the size, number of bedrooms,

Â the zip code or postal code, and the wealth of the neighborhood.

Â And so given these input features,

Â the job of the neural network will be to predict the price y.

Â And notice also that each of these circle, these are called hidden units in

Â the neural network, that each of them takes its inputs all four input features.

Â So for example, rather than saying these first nodes represent family size and

Â family size depends only on the features X1 and X2.

Â Instead, we're going to say, well neural network,

Â you decide whatever you want this known to be.

Â And we'll give you all four of the features to complete whatever you want.

Â So we say that layers that this is input layer and

Â this layer in the middle of the neural network are density connected.

Â Because every input feature is connected to every one

Â of these circles in the middle.

Â And the remarkable thing about neural networks is that, given enough data about

Â x and y, given enough training examples with both x and y, neural networks

Â are remarkably good at figuring out functions that accurately map from x to y.

Â 6:48

So, that's a basic neural network.

Â In turns out that as you build out your own neural networks,

Â you probably find them to be most useful, most powerful

Â in supervised learning incentives, meaning that you're trying to take an input x and

Â map it to some output y, like we just saw in the housing price prediction example.

Â In the next video let's go over some more examples of supervised learning and

Â some examples of where you might find your networks to be incredibly helpful for

Â your applications as well.

Â