Jens Sjölund

Hi @arokem , regarding your performance question: the probabilistic least-squares version will probably come with a small performance hit, although I haven't timed it. I guess you could have an...

Hi! I'm glad you picked this up @Garyfallidis! I'll rebase as soon as possible (just have to refresh my memory first).

Hi again! I hope it is correctly rebased now.

I thought it would be possible to fix this by making sure the initial guess was zero for the deactivated axes (because the gradients should be zero). But apparently I...