How to train the dataset more than once?

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

How to train the dataset more than once?

manal alghamdi


Dear Wekalist,

I have heard that sometimes the more we trained the dataset the better the performance.  I really do not know other way than downloading the dataset and choose the classifier and press the button "Start"; then, we get the result.  Is there any other way to train the dataset more in WEKA using the GUI?

Thank you so much

Manal Alghamdi

_______________________________________________
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: How to train the dataset more than once?

Eibe Frank-2
Administrator
There are two general types of classifiers that can be trained “more”:

* Classifiers that implement the IterativeClassifier interface (http://weka.sourceforge.net/doc.stable-3-8/weka/classifiers/IterativeClassifier.html). These classifiers are currently all variants of boosting, but some other classifiers such as MultilayerPerceptron, etc., should be added (they implement the basic iterative functionality, but do not yet implement the IterativeClassifier interface).

* Classifiers that implement the UpdateableClassifier interface (http://weka.sourceforge.net/doc.stable-3-8/weka/classifiers/UpdateableClassifier.html).

In an IterativeClassifier, the number of iterations determines how long the classifier is trained (e.g., how many base classifiers are added to the ensemble classifier). You can use the IterativeClassifierOptimizer to automatically select an appropriate number of iterations using internal cross-validation.

In an UpdateableClassifier, you will need to write some code or script to run the training data through the learning algorithm multiple times. I don’t think there is an automatic way to do this in WEKA (but I might be wrong, perhaps it can be done with the KnowledgeFlow).

Overfitting is obviously an issue in both cases, so internal cross-validation (or similar) should be used to decide when to strop training.

Cheers,
Eibe

> On 3/01/2017, at 6:52 PM, manal alghamdi <[hidden email]> wrote:
>
>
>
> Dear Wekalist,
>
> I have heard that sometimes the more we trained the dataset the better the performance.  I really do not know other way than downloading the dataset and choose the classifier and press the button "Start"; then, we get the result.  Is there any other way to train the dataset more in WEKA using the GUI?
>
> Thank you so much
>
> Manal Alghamdi
> _______________________________________________
> 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
Loading...