2011/12/7 Gael Varoquaux <[email protected]>: > On Tue, Dec 06, 2011 at 10:26:04AM -0500, Ian Goodfellow wrote: >> I agree with David that it seems like the optimizer is broken, but I >> disagree that the problem is the termination criterion. There should >> not be any NaNs anywhere in the course of optimization. > > I have also seen problems with NaNs in the case of very correlated > dictionnaries/designs. > > I think that this is due to the conditionning of the problem, and > numerical erros that accumulate. I looked a bit, but really couldn't find > where the problem was. I had too leave for travels, but Virgile, Fabian > and I where hoping to work on this more. > > Can people confirm that some other solvers (e.g. GLMnet) do not have the > same problem? In which case, we need to figure out how they do it.
Are we still talking about the Lasso with Coordinate Descent or another estimator? Would be great to have a reproduction case in a gist and a matching issue ticket on github not to forget about it. -- Olivier http://twitter.com/ogrisel - http://github.com/ogrisel ------------------------------------------------------------------------------ Cloud Services Checklist: Pricing and Packaging Optimization This white paper is intended to serve as a reference, checklist and point of discussion for anyone considering optimizing the pricing and packaging model of a cloud services business. Read Now! http://www.accelacomm.com/jaw/sfnl/114/51491232/ _______________________________________________ Scikit-learn-general mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
