complete separation - convergence issue in GLMs - lassoGLM

2 views (last 30 days)
How can I deal with convergence issue in GLM?
I tried to use penalized logistic regression. But instead of plots like in the documentation lassoglm
I get the plots below (code also below). Looks like lambda should be close to zero.
I guess this is the convergence issue I hoped to avoid using penalized GLM. My question: why does penalization here not work and how can I make working it?
Plot 1:
Plot 2:
load('GroundTruth.mat') % attached to question
Ybool = label;
X = features;
[B,FitInfo] = lassoglm(X,Ybool, 'binomial',...
'NumLambda',25,'CV',10);
% figure
lassoPlot(B,FitInfo,'PlotType','CV');
% figure
lassoPlot(B,FitInfo,'PlotType','Lambda','XScale','log')

Answers (0)

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!