How does 'prior' and 'cost' parameters influence esembling learning ?

I am interested in how 'prior' and 'cost' are implemented in esembling learning algorithm of MATLAB. More specifically, how they are implemented in 'AdaboostM1' and 'Bag'.

1. How does they influence the training process at ensemble level ?
2. How does they influence the training process at weaker learner level ?

Answer 1

[Hiro Yoshino]

This link seems to have an answer for it together with other links to relevant information.

[Keyan Li]

Hi Hiro, Thanks for your reply. I have looked at that page. I am more interested to know more details about the implementation. For example, in this page, it explains how observation weights influence the training process of "bag" and "AdaboostM1". I am interested in the implications of "prior" and "cost" in "bag" and "AdaboostM1" as well.