underlying data model, the underlying process which generated the data.

But the other point to look out for

here is it's sensitive to perturbations in the data.

If I move just one of these points a little bit,

how much does that curve change?

And the green one would not change that much while the red one might.

Okay.

But, I actually don't think this curve fitting model is the best image to call up

when you think about overfitting because in my mind, it's a little bit specific to

polynomial curves and the number of degrees of freedom you're working with.

Okay.

So it's not always clear, at least to me, how to map this image that I call of my

head of over fitting to machine-learning problems.

So I think a more useful one is one that's actually associated with

a Wikipedia page on an article related to overfitting that's

released under Creative Commons is this, that over time,

your error goes down for both your training set and your test set.

But at some point it continues to go down for your training set.

I'm pointing at the wrong screen.

This continues to go down on your training set but

then it starts to creep up on your test set.

So, at this point,

he's put a little symbol here to indicate that's where overfitting has set in.

And so I think this is the images that you call to mind when you're thinking about

overfitting problems.

This is the difference between the error on your training test and

the error on a test set.

Now, in some cases it's tough because

the test set may be, what is the test set?

Is it something you actually have you hands on where you can measure this error,

or is it more like predictive power in the real world?

But regardless, any kind of estimate you have over the error that you're

actually achieving on data that you didn't train on is what to look out for.

When these start to diverge, that means you overfit.

All right.

So, other language to be familiar with around this concept,

is the model able to generalize?

If it is able to generalize then you are not overfit.

Okay. Can it deal with unseen data or

does it overfit the data it test on?

So in order to solve for this, you test on hold-out data.

Okay.

So one way to do this is split the data to be modeled in the training test set,

just in a fixed way.

Train the model on the training set.

Evaluate the model on the training set, and evaluate the model on the test set.

And the difference between those two,

again, is a measure of the model's ability to generalize.

A measure of how overfit it is.

Now, doing this just once, splitting the training and test,

is not the most powerful mechanism to do this,

and in a couple of slides I'll give you a better one.