Contents
How many scaling can be applied in SVM?
I think there are two scaling can be applied. scale each instance vector such that each vector has zero mean and unit variance. According to my experiments with RBF kernel (libSVM) I found that the second scaling (2) improves the results by about 10%.
Which is a property of the SVM classifier?
Training an SVM classifier includes deciding on a decision boundary between classes. This boundary is known to have the maximum distance from the nearest point on each data class. Due to this property, SVM is also referred to as a maximum-margin classifier:
How is the geometric margin defined in SVM?
The geometric margin γ̂ is defined as the distance of the i-th observation to the decision boundary. Unlike the functional margin, this measure is invariant to the scaling of parameters. After all, the hyperplane defined by w ᵀ x + b = 0 is exactly the same as the one defined by 10 w ᵀ x + 10b = 0.
How does the SVM choose its line [ video ]?
Here we show two decision boundaries represented by line 1 and 2. It is easy to see that there are many other lines that can divide the input space in a way that all observations are correctly classified — in fact, there are an infinite number of lines. So how does the SVM choose its line? There are two main ideas.
Why do we need to scaling kernel values?
Because kernel values usually depend on the inner products of feature vectors, e.g. the linear kernel and the polynomial ker- nel, large attribute values might cause numerical problems. We recommend linearly scaling each attribute to the range [-1,+1] or [0,1]. I believe that it comes down to your original data a lot.
When do you lose definition when scaling linearly?
If your original data has SOME extreme values for some columns, then in my opinion you lose some definition when scaling linearly, for example in the range [-1,1]. Let’s say that you have a column where 90% of values are between 100-500 and in the remaining 10% the values are as low as -2000 and as high as +2500.