What is the meaning of gamma on Svm
"gamma" is a Hyperparameter in svm that is used to control the radius of the area of influence of the support vectors. YOu can see from the formula as
K(xi,xj)= exp(-gamma|| xi - xj || ^2 ), gamma>0.
It is a Hyperparameter for controlling this K nonlinear kernel function called "Gaussian radial basis function."
You can clearly see by the mathematical relation that.
- If gamma is small then K will be large means it will have high influence on deciding the class of the vector x_i.
- If gamma is large, it will have less influence on deciding the class of the vector x_i.
In short : Large gamma leads to high bias and low variance models, and vice-versa.