MAchine Learning Models.

I have a very basic questions

1. What is the number of trees to use for a RF model in R and  the PP Tree models.  For Molecular datasets with 100 mols, 500 mols, 1000 mols, 5000 mols , 10000 mols, 100000 mols

2. What are the optimal choices fro SVM.

3. How to use the inbuilt PLS and KNN methods in an optimal manner. Some key learnings might help.

4. What is the fork error for the R methods in PP 9.0

A wrong answer is always better than no answer.

Some nice discussion on the Machine learning methods in PP 9.0 might be helpful. I am expecting Dana Honeycutt to write a few lines in response to this request.

Best

Subhas