Levenberg-Marquardt.NET

Download LMA.NET (.NET 6.0 / Nunit 3.0)

The link above contains the full source code, including NUnit test files and VStudio 2013 solution of the LMA.NET library. The algorithm has been updated on 6/25/2010 - there was a defect due to low value of delta in the derivative function that sometimes caused singularities. The new value of delta is 1*10-9 which seems to work now.

Public constructor: public LMA(LMAFunction function, double[] parameters, double[][] dataPoints, double[] weights, GeneralMatrix alpha, double argDeltaChi2, int argMaxIter)

• function: function to fit the experimental data. Inherits from the LMA Function
• parameters: vector with initital parameters
• dataPoints: array containing experimental data that will be used in the fit
• weights: array of doubles with weights to experimental points
• Alpha: matrix
• epsilon: accuracy of the final result
• itMax: max number of iteration

function to call to execute minimum search: public void Fit()

Mimium is stored in the Parameters property of the object