Open KnutJaegersberg opened 2 years ago
You gotta know, your naive bayes model scales really well record and feature wise, as long as you have enough ram. Can I use it on a lot of classes, too, out of the box?
(I had cases where I considered using vowpal wabbit, but could actually just fit the problem in (128 gb) RAM with textmodels nb and your dfm implementation, great work)
Is there a maximum number of accepted factor levels when training naive bayes? I just only got missing values returned on a multiclass classification problem with 37 factor levels and 25k records.