Adaboost

classic Classic list List threaded Threaded
4 messages Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Adaboost

Riccardo Granero
Dear all,

only a confirm!

After the training phase, for classifying unlabeled test istance Adaboost use the weighted (from the training phase) magiority voting?

Thanks
Riccardo
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Adaboost

Marcus Müller
Hello Riccardo,

AdaBoost is a so called ensemble method, utilizing so called base learners, like Decision Tree , kNN, SVM etc.
The AdaBoost trains multiple of those base learners and comes to a decision by averaging (regression) or a majority vote (clasifiaction) of the base learner's results. As you already mentioned this happenes in a weighted fashion as the "importance" of additional base learners decreases with their number. In AdaBoost the k+1 base learner tries to get those samples right which have been wrongly classified by the previous kth base learner. The training data for both base learners is the same, with the only difference, that the weights of those wrongly classified samples in the kth base learner are increased for base learner k+1. This is repeated until a desired performance or a specified number of base learners is reached.

Best regards,
Marcus

> Am 12.06.2017 um 10:57 schrieb Riccardo Granero <[hidden email]>:
>
> Dear all,
>
> only a confirm!
>
> After the training phase, for classifying unlabeled test istance Adaboost
> use the weighted (from the training phase) magiority voting?
>
> Thanks
> Riccardo
>
>
>
> --
> View this message in context: http://weka.8497.n7.nabble.com/Adaboost-tp40921.html
> Sent from the WEKA mailing list archive at Nabble.com.
> _______________________________________________
> Wekalist mailing list
> Send posts to: [hidden email]
> List info and subscription status: https://list.waikato.ac.nz/mailman/listinfo/wekalist
> List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
List info and subscription status: https://list.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Adaboost

Eibe Frank-2
Administrator
Yes, it's a weighted vote. The weight of a classifier in the ensemble is based on its error rate on the weighted data it was trained on. Check the original paper for details.

Cheers,
Eibe

> On 12/06/2017, at 9:33 PM, Marcus Müller <[hidden email]> wrote:
>
> Hello Riccardo,
>
> AdaBoost is a so called ensemble method, utilizing so called base learners, like Decision Tree , kNN, SVM etc.
> The AdaBoost trains multiple of those base learners and comes to a decision by averaging (regression) or a majority vote (clasifiaction) of the base learner's results. As you already mentioned this happenes in a weighted fashion as the "importance" of additional base learners decreases with their number. In AdaBoost the k+1 base learner tries to get those samples right which have been wrongly classified by the previous kth base learner. The training data for both base learners is the same, with the only difference, that the weights of those wrongly classified samples in the kth base learner are increased for base learner k+1. This is repeated until a desired performance or a specified number of base learners is reached.
>
> Best regards,
> Marcus
>
>> Am 12.06.2017 um 10:57 schrieb Riccardo Granero <[hidden email]>:
>>
>> Dear all,
>>
>> only a confirm!
>>
>> After the training phase, for classifying unlabeled test istance Adaboost
>> use the weighted (from the training phase) magiority voting?
>>
>> Thanks
>> Riccardo
>>
>>
>>
>> --
>> View this message in context: http://weka.8497.n7.nabble.com/Adaboost-tp40921.html
>> Sent from the WEKA mailing list archive at Nabble.com.
>> _______________________________________________
>> Wekalist mailing list
>> Send posts to: [hidden email]
>> List info and subscription status: https://list.waikato.ac.nz/mailman/listinfo/wekalist
>> List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
> _______________________________________________
> Wekalist mailing list
> Send posts to: [hidden email]
> List info and subscription status: https://list.waikato.ac.nz/mailman/listinfo/wekalist
> List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
List info and subscription status: https://list.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Adaboost

Riccardo Granero
Ok, thanks a lot!

Riccardo
Loading...