SVR: Support Vector Regression

Question 1

What distinguishes SVR from linear regression?

Answer 1

Instead of the model being a basic line

It is a basic line with a tube around it:
The (epsilon) Ɛ-Insensitive Tube

~~~~~~~~~~~~~~~ Ɛ tube 
                 x
————————- line 
x
           x

~~~

Question 2

What’s the point of the Ɛ-Insensitive tube?

Answer 2

It allows our data to have a margin of error

The distance (error) between any data point inside the tube and the line doesn’t matter

             x  Ɛ(small)4
      x      :

x. x
—————————-
x
~~~~~~~~~~~~~~~~
:
x Ɛ1* (ps small 1) * = below tube
~~~

We’re calculating the distance between the data points outside the tube and the tube

The points outside the tube dictate it’s orientation

Question 3

Why’s it called support vector regression?

Answer 3

Because the data points are vectors

And the tube’s direction is dictated by the data points (vectors) outside of it

_____________
x (vector)
~~~~~~~~~~~
:
x (support vector)
~~~

So they’re referred to as support vectors