RBF-Network

Python implementation of Radial Basis Function network with no external packages

Sample data setup

Function defination:

where x ranges from 0~1 and the noise ranges from -1 to 1

The final curve looks like this:

Training process

The training process includes forward-pass, backward-pass and online update for the weights. Using k-means to dertermine the centers and variance of the training data and use these two parameters as the initial setup

Result