Hi Yin,
I figured out how to do a proper pull request - here it is.
I've added a few more modifications and it is safe to merge my changes now. By default, missing values are not accepted - if they are present, a message "missing values not allowed" shows up and the code proceeds to crash exactly as it would have done before. In order for missing values to be converted to NaNs, one needs to specifically call InstanceReader.read(...) with an additional flag argument set to true.
Never clause is simply parsed properly now - no side effects, doesn't depend on predictor type. Safe to merge. I haven't changed any code for the actual predictors, they all should work as before.
Let me know if you have any more concerns - I can do further tests or modifications if you want. Otherwise I would appreciate if you approve the pull request before introducing any new changes to mltk.
Hi Yin, I figured out how to do a proper pull request - here it is.
I've added a few more modifications and it is safe to merge my changes now. By default, missing values are not accepted - if they are present, a message "missing values not allowed" shows up and the code proceeds to crash exactly as it would have done before. In order for missing values to be converted to NaNs, one needs to specifically call InstanceReader.read(...) with an additional flag argument set to true.
Never clause is simply parsed properly now - no side effects, doesn't depend on predictor type. Safe to merge. I haven't changed any code for the actual predictors, they all should work as before.
Let me know if you have any more concerns - I can do further tests or modifications if you want. Otherwise I would appreciate if you approve the pull request before introducing any new changes to mltk.