Support Vector Machine

SVM学习笔记

1. The target of SVM is finding a hyper plane to make margin max.

2. How to get Max Margin Hyper plane?

The distance of hyper plane and the closest plots of two sides is equal.

3. Linear separable and Linear inseparable

Linear separable

Linear inseparable

4: Hyper plane

W*X+b=0 W={w1,w2,w3...wn}

W: weight vector

X: training instances 

n: number of features

b: bias

5. X=(x1,x2)

right top plot: 

left down plot:

Adjust weight:

we get 

That means whatever y=1 or -1, w0+w1x1+w2x2 always more than 1.

support vectors is the max margin hyper plane on both sides.

The distance of one side and the middle hyper plane is 1/||w||, so the max distance is 2/||w||.

||w|| is the norm of vector.

5. How the SVM find the MMH?

L is the number of the support vector plots, yi is the support vector plots, xi is the class label, Xt is the testing instance.

ai and bi is the single value.

We can use this formula to get a value,  according to the value is negative or positive, we could classify it.