Some Classes are not Linearly Separable
It's pretty easy to imagine classes that are not linearly
separable.
What would you do?
There are lots of things you could do.
You could do several lines.
One popular solution is the kernel trick, which we will discuss
in the Support Vector Machine
lecture.
Another question is outliers. If you classify 99% of the data
correctly, do you care about the remaining bit?
What would you do if your classifier could be less than perfect?
