A good theoretical background of this subject can be found here.
This site makes available an online demonstration of the regressor, as shown in the figure on the right, to better grasp its concepts. First, click on the "Add sample point", and then on the "Plot for these values". Get the feeling of the alfa parameter by changing it; lower alfas tend to overfit, while large alfas tend to smooth and eventually overfit. The beta parameter is used for the other kernels, but they are far less efficient than the Gaussian kernel.
This application requires .NET 3.5 and you MUST USE IE7.
If you do not have it, you can freely update your computer.
If you are all set, you can fire up the online demostration bu clicking here!