Radial basis function networks

Radial Basis Function Networks

• Global approximation to target function, in terms of linear combination of local approximations
• Used e.g. for image classification
• A different kind of neural network
• Closely related to distance-weighted regression, but “eager” instead of “lazy”

Training Radial Basis Function Net work

Q1: What x_u to use for each kernel function K_u(d(x_u, x))

• Scatter uniformly throughout instance space
• Or use training instances reects instance distribution)

Q2: How to train weights (assume here Gaussian K_u)

• First choose variance (and perhaps mean) for each K_u

-e.G.- use EM

• Then hold K_u fixed and train linear output layer

-efficient methods to t linear

RBF use three functions: