Regularization breaks LogisticRegressionAnalysis in 2.14.0

[marklam]

I have a set of data which I could calculate a logistic regression for in 2.13.1, but not in 2.14.0.
I still get a 'converged' result from Compute, but the coefficients are all -1.2748188443139669E+194.
If I use the same procedure used in LogisticRegressionAnalysis, but use an IterativeReweightedLeastSquares whose Regularization is set to 0, everything works as in 2.13.1

I've uploaded the data that shows the problem

[cesarsouza]

Thank you! I will write a unit test to reproduce the issue and investigate.

[cesarsouza]

Hi there, I have fixed the regularization implementation and added a new property in the LogisticAnalysis class to manually override the regularization options if needed. I will be available in the next release but I can also generate a temporary alpha release if there is interest.

Hope it helps!

Best regards,
Cesar

[cesarsouza]

A new pre-release NuGet package has been created with the fix. Thanks for reporting the issue!