RBF Network Fitting is an in Python developed fitting routine, which is using the Radial-Basis-Function-Network for solving the 1D- and 2D-minimization problem. During the Self-Consistent-Field-Optimization of the RBF-Network, the mean-squared-error will be evaluated for each cycle, and a difference- and gradient-correction will be applied to the input-parameter of the Fitting-Model. As Fitting-Models can be choosen:

• Normal Distribution
• Cauchy/Lorentzian Distribution
• Pseudo-Voigt Profile

In order to optimize the Hyperparameter-Finding for the number of layers and the kind of choosen models, a Genetic Algorithm can be optionally used. The combination of both Radial-Basis-Function-Network and Genetic Algorithm allows using RBF Network Fitting as a real black-box-method in the absence of empirical parameters.

• Detecting peaks of an oscillating function

Example - I	Example - II

• Fitting of experimental data

Example - III

• Following patterns of 3D-Functions

Example - IV	Example - V

RBF Network Fitting requires:

• numpy
• matplotlib

Installing and Running:

python setup.py install
# as command line application 
python -m RBFN 
# as library
from RBFN import GeneticFitter
from RBFN import RBFNetwork
from RBFN import PlotResults

Genetic Algorithms and Machine Learning for Programmers: Create AI Models and Evolve Solutions
Frances Buontempo
Pragmatic Bookshelf, 2019

Genetic Algorithms with Python
Clinton Sheppard
Clinton Sheppard, 2018

https://github.com/handcraftsman/GeneticAlgorithmsWithPython/blob/master/ch08/genetic.py https://en.wikipedia.org/wiki/Radial_basis_function_network