bobaronoff

I must not be conveying the issue correctly. Here is my understanding and working with my data bears out that understanding. ```unsafe_wrap``` takes the C pointer and uses it to...

Thank you!

I apologize. This is third submission. I am unable to figure out how to squash my commits in order for them to be reviewed. This is my first PR ever.

I will work on a new version - should not take long. I do not understand how to submit a revised PR as opposed to creating a new PR. Am...

I was able to make some modifications. They are showing above so I guess I don't need to do anything special to revise PR. ```margin``` is re-added. Default for ```type```...

I think it best if we leave the current 'predict' alone and rename my version , I chose 'predictbytype'. It avoids all the issues you correctly point out. Altering 'type'...

Following up on your reply ( although don't see above). My thinking is that having a function return three different types makes a complex scenario even worse. 90+ % of...

From past work in R, early stopping is a convenience feature provided by the package as opposed to a libxgboost function. The feature relies on capturing the evaluation log and...

Not sure I understand. I am able to set the evaluation metrics with the eval_metric keyword parameter. The results show up in the log and are parse-'able. It does require...

Recopied from #148. This is my function to capture the logs. ``` function xgboost_log(traindm::DMatrix, a...; testdm::Any=[] , num_round::Integer=10, kw... ) Xy = XGBoost.DMatrix(traindm) b = XGBoost.Booster(Xy; kw...) update_feature_names::Bool=false if typeof(testdm)==DMatrix...