I want to know whether there are works on empirical comparison between two multi-class AdaBoost algorithms, AdaBoost.MM proposed in the paper "A Theory of Multiclass Boosting" (section 8) and SAMME/SAMME.R proposed in "Multi-class AdaBoost".
The first paper "A Theory of Multiclass Boosting" did some theoretical analysis of SAMME, but there are no comparison of these two methods in the numerical experiment section of the paper. I did a literature research and didn't find any useful result.
I can code both of them up and test them on datasets from, say, UCI machine learning repository, but experiments on larger and not very commonly used datasets arising from real-world problems, are more appreciated.
