Kernelized SVM.txt

Kernelised SVM
Gamma controls how far the influence of a single training model reaches
Gamma- small value of gamma- simpler models, smoother decision boundaries, even points further apart are considered similar, results in more points being grouped together and smoother decision boundaries.
small gamma-larger similarity radius.

Larger values of gamma- Kernel value decays quickly and points have to be close to be considered similar.
This results in more complex, tightly constrained decision boundaries.

Just like in linear SVMs, with kernelised SVMs we can also vary C,which interacts with gamma.
If gamma is large then C will have little or no effect.
If gamma is small, the effect of C will be the same as with a linear classifier.