[MUSIC] This course address classification,
which is one of the most widely used, most fundamental areas
of machine learning. If you understand classifiers, you'll also understand basically the rest
of machine learning and the techniques we use here is what most people in
the industry need to be successful. We discuss how much learning
is about input data which can be pushed through
some machinery algorithm which outputs what we think about this
kind of intelligence from the data. In this course,
we're going to build classifiers, so classifier takes into input But
some x or some features of our data. And as output makes a prediction
which a discrete class or category or label for the data. And we're going to see a ton of different
examples of how this is used for in practice. The goal of a classifier is to learn a
mapping from the input x to the output y, those classes. The example that we discussed in the first
course was a sentiment classifier, where we're given that input sentence x,
like easily the best sushi in Seattle. We fed that through
the sentiment classifier, which then told us an output y. That was either,
yeah that is a positive sentence. Or nay, that is a negative sentence. And we can use these sentences,
these predictions in a wide range of ways, as we'll see soon. A general classifier about is about,
taking some input x, pushing it through some model,
which predicts why that might be, for example, two classes or multiple classes
of, say, positive or negative, or as we will see, it could be three,
four or more categories. Let's suppose for
example I have a web page, and I want to figure out what ads
to show on this web page. So I need to figure out what
this web page is about. The goal here is to take the text of the
web page and categorize it automatically as to whether it's kind of an educational
site and educational type ads. Whether it's kind of a site about
finance or an article about finance and we need that kind of ad. Or one about technology and so on. So a classification is not jut binary,
positive or negative, but it can be one of multiple categories,
or multiple classes. Perhaps the most common type of
classifier that we see everyday, every time we open up our
email is a famous spam filter. So the spam filter takes
every time an email arrives. Makes a prediction whether this is a spam
email, it should be ignored or not spam. And that prediction needs to be made based
not just on the text of the email but on other information we
get from that email. Like who that sender was what the IP
address of that sent message is. Other messages that the sender sent, and
so on, and, from that information, we're going to learn the mapping, from those
inputs, to whether it's spam, or not. And, those spam filters gotten so
much better over the years. So, I remember, early on,
we just used keyword search, or keyword classifiers, and
they weren't very good. But, today I don't even check
my spam folder anymore. So if you send me an email and didn't
open it, maybe it's in my spam folder. Sorry. We can build all sorts
of classifiers though. We can use, for example, image data. So given this particular input,
my dog, the image pixels. I want to make a prediction
from a certain category. So from this famous image net data set, where there's a thousand different
categories you might want to predict. So for example, if you want to know if it's a labrador
retriever, a golden retriever, and so on. What kind of dog it is, and that's
the output label y that we might want. Now the idea of classifiers
can be extremely useful for a wide range of domains. One that I'm particularly excited about
is the area of personalized medicine, which I think is going to
change the world. So today if I don't feel so well I might
put a thermometer under my arm and check out my temperature,
or a doctor might order an X-Ray to see what's going on in my
chest, or maybe use some lab tests. And based on that information,
goes through some crossfire, which maybe it's in the doctor's head or
maybe it's an automated system, that tries to make a prediction as
to what condition I might have. But what's annoying about
how medicine is done today is that based on the same conditions,
we make the same predictions for me or for you independent of the fact that
we're really different people. Personalized medicine aims
to totally change that. So it's going to look at our DNA sequences
because we're genetically different and find a good treatment for each one of us. And maybe even look at our lifestyle
which might say something about what I'm prone to. Maybe that's your lifestyle,
maybe my lifestyle is more like this. And so based on that kind of information,
we can predict what condition I have and what have treatment is going to
be the most effective for me. And that's an example of
classification in the real world. Perhaps one of the most fun and
surprising examples of classification is work that one of my colleagues
Tom Mitchell did, which is pretty amazing. You take a scan of your brain
as you look at a word, and based on that image from
what's called an FMRI, he can make a prediction as to
what kind of word you're reading. So for example, based on the image of your
brain, it can predict if you're reading, say, the word hammer or the word house. Which is basically reading your mind. And I've been talking to Tom for
a long time about this topic. More than ten, 15 years. And over that time,
the concept results they have, had evolved from very basic things. So amazing things. So, for example today, they can train a classifier on your brain
images based on words that you read, and then use it to predict something from my
brain images based on pictures that I see. So picture of a hammer instead
of the actual word hammer. And that is an incredible
kind of evolution, an incredible kind of analysis
that you can do from brain data. Really, really cool
example classification. Read your mind. [MUSIC]