WEBVTT

00:00:00.280 --> 00:00:03.950
Now I want to show you a visualization tool that I found online that I

00:00:03.950 --> 00:00:09.350
think does a really great job of helping you see what k-means clustering does.

00:00:09.350 --> 00:00:12.160
And that should give you a good intuition for how it works.

00:00:12.160 --> 00:00:15.800
So I'd like to give a special shout out to Naftali Harris,

00:00:15.800 --> 00:00:19.890
who wrote this visualization and very kindly agreed to let us use it.

00:00:19.890 --> 00:00:22.990
I'll put a link to this website in the instructor notes that you can go and

00:00:22.990 --> 00:00:24.640
play around with it on your own.

00:00:24.640 --> 00:00:29.720
So it starts out by asking me how to pick the initial centroids of my clusters.

00:00:29.720 --> 00:00:32.530
I'll start out with Randomly right now.

00:00:32.530 --> 00:00:34.490
What kind of data would I like to use?

00:00:34.490 --> 00:00:35.800
There are a number of different things here, and

00:00:35.800 --> 00:00:37.500
I encourage you to play around with them.

00:00:37.500 --> 00:00:40.630
A Gaussian Mixture has been really similar to one of the simple examples we've

00:00:40.630 --> 00:00:41.670
done so far.

00:00:41.670 --> 00:00:43.490
So Gaussian mixture data looks like this.

00:00:43.490 --> 00:00:45.460
These are all the points that we have to classify.

00:00:45.460 --> 00:00:46.540
The first question for you is,

00:00:46.540 --> 00:00:49.470
how many centroids do you think is the correct number of centroids on this data?