Artur Skrzeta

Support Vector Machine with Numpy and Python

Intro

• Splitting the data in the best posible way - to get two data groups (categories).
• The key aspect is to split the data in as much optimal way as possible so that new data point can be precisely clasified.
• The data can be split in different ways. However, the line making the biggest gap between two classes is chosen as the separation line.

• Support vectors are extreme points of a specific data set of one category.
• Support vectors are those elements of the data that makes the biggest possible separation gap.
• Distance margin is the distance between support vectors from both data classes.
• Once we create an optimal separation line we can easily see which side a new data belongs to.
• The smaller the distance margin, the higher a chance of miss-classification is.
• If all of the data is placed in one line in 1D then we need to convert it into 2D with kernel.
• If one data of one class is sourrounded by data of second class we need to convert 2D into 3D with kernel.