Adaboost

[Naoki Wake]

Hi

I am learning about boosting and testing adaboost using milk.

I referenced the following page 

http://packages.python.org/milk/adaboost.html?highlight=ada#module-milk.supervised.adaboost

and noticed that this example-code creates only one weak learner.

Adaboost uses some weak learners, so I want to know how to combine some weak learners and create a strong learner.

[Luis Pedro Coelho]

Hi,

In milk's terminology, a learner is an object which can be called upon to
generate a classifier (what milk calls a model).

Therefore, the code will repeatedly use the learner you pass it to generate
multiple models. So, this will do what you expect.

HTH,
Luis

--
Luis Pedro Coelho | Institute for Molecular Medicine | http://luispedro.org

[Naoki Wake]

Luis,

Thank you for reply.

 I understood that the code generate multiple models.

Then, can I set the number of models which are combined?

 

Naoki

[Luis Pedro Coelho]

On Thursday, October 11, 2012 09:39:37 AM Naoki Wake wrote:
> Luis,
> Thank you for reply.
> I understood that the code generate multiple models.
> Then, can I set the number of models which are combined?

You can set the max_iters parameter:

learner = milk.supervised.adaboost.boost_learner(weak, max_iters=8)

This is the maximum number of weak learners, but it might be less if the
algorithm believes it is no longer making progress.

HTH

[Naoki Wake]

Luis,

I understood how to do that, thank you for your kind reply.

Naoki