Generalized Additive Model

Similar to GLM but we allow any weight function (can be nonlinear too) $g\left(E_Y(y\mid x)\right)={\beta }_0+f_1\left(x_1\right)+f_2\left(x_2\right)+\dots +f_p\left(x_p\right)$

So we add a set of basis functions for each feature as new features.

• Spline
• Radial Basis Function Kernel
• Shifted Sigmoid Function