Closed huroh closed 8 years ago
If I cut down to around ~7 columns it works.
I've noticed I also get memory leaks when running CREATE MODEL when I use 'NA' instead of 'nan' to indicate missing values, in a column that is otherwise numerical. Let me know if you need an example for this.
Again, thanks for helping us find all these bugs. I am not sure we explicitly support 'NA' to indicate missing values right now, but we should definitely add support for that.
Also, I am curious: what is the data you're trying to analyze? Is it some type of psychology experiment with a variety of tasks, and how well each subject did on each task?
Not psychology, it's educational analytics, the aim is early prediction of student success. I'm interested in comparing the methods in bayesdb with various machine learning models I've already been using.
[Using VM provided]
CREATE BTABLE training FROM data.txt; CREATE 10 MODELS FOR training;
https://dl.dropboxusercontent.com/u/68514/bayesdb/data.txt