Introduction

##Supervised Classification

• We know the right answer to problems and the algorithm learns by examples.

• Take in features as input and give out labels as outputs.

• Decision Surface - Boundary (can be more than two dimesional) which sepeartes different samples belonging to different classes. Is made by learning from the given data and then can be used to seperate unknown samples into classes according to their features.

• Good Decision Surface - correctly classifies (as much as possible) for the genral case and is robust.
• Robustness - A decision surface is more robust if a small error in our test data will not lead to wrong classification. In figure below the robust linear surface is checked. The other one is too close to red crosses.