Shrinkage value above 1 AMT

classic Classic list List threaded Threaded
5 messages Options
Reply | Threaded
Open this post in threaded view
|

Shrinkage value above 1 AMT

bujarra
Hi,

Does it make sense to set the shrinkage value above 1 in AMT (I even get
better results with a value of 1.1-1.4?. Which is the unit of this
parameter?



--
Sent from: https://weka.8497.n7.nabble.com/
_______________________________________________
Wekalist mailing list -- [hidden email]
Send posts to [hidden email]
To unsubscribe send an email to [hidden email]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/postorius/lists/wekalist.list.waikato.ac.nz
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
Reply | Threaded
Open this post in threaded view
|

Re: Shrinkage value above 1 AMT

Eibe Frank-3
Interesting. Shrinkage values > 1 don't immediately make sense to me. They imply that every time a linear regression model has been fitted to the residuals remaining from the previous iteration, this model's predictions are multiplied by a value > 1 before the residuals to be fitted by the linear model(s) in the next iteration of "boosting" are computed.

Cheers,
Eibe

On Fri, Apr 30, 2021 at 11:22 AM bujarra <[hidden email]> wrote:
Hi,

Does it make sense to set the shrinkage value above 1 in AMT (I even get
better results with a value of 1.1-1.4?. Which is the unit of this
parameter?



--
Sent from: https://weka.8497.n7.nabble.com/
_______________________________________________
Wekalist mailing list -- [hidden email]
Send posts to [hidden email]
To unsubscribe send an email to [hidden email]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/postorius/lists/wekalist.list.waikato.ac.nz
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html

_______________________________________________
Wekalist mailing list -- [hidden email]
Send posts to [hidden email]
To unsubscribe send an email to [hidden email]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/postorius/lists/wekalist.list.waikato.ac.nz
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
Reply | Threaded
Open this post in threaded view
|

Re: Shrinkage value above 1 AMT

bujarra
Although it isn't usual, it isn't wrong, right? However, do you recommed to
keep the value fix to 1? I see a variety of results by modifying this
parameter.



--
Sent from: https://weka.8497.n7.nabble.com/
_______________________________________________
Wekalist mailing list -- [hidden email]
Send posts to [hidden email]
To unsubscribe send an email to [hidden email]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/postorius/lists/wekalist.list.waikato.ac.nz
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
Reply | Threaded
Open this post in threaded view
|

Re: Shrinkage value above 1 AMT

Eibe Frank-3
How do you measure predictive performance? Using RMSE? Are the differences in performance statistically significant?

Cheers,
Eibe

On Sun, May 2, 2021 at 8:43 PM bujarra <[hidden email]> wrote:
Although it isn't usual, it isn't wrong, right? However, do you recommed to
keep the value fix to 1? I see a variety of results by modifying this
parameter.



--
Sent from: https://weka.8497.n7.nabble.com/
_______________________________________________
Wekalist mailing list -- [hidden email]
Send posts to [hidden email]
To unsubscribe send an email to [hidden email]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/postorius/lists/wekalist.list.waikato.ac.nz
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html

_______________________________________________
Wekalist mailing list -- [hidden email]
Send posts to [hidden email]
To unsubscribe send an email to [hidden email]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/postorius/lists/wekalist.list.waikato.ac.nz
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
Reply | Threaded
Open this post in threaded view
|

Re: Shrinkage value above 1 AMT

bujarra
Being Cartesian coordinates my unique two attributes, I check the lineal
distance between original and predicted value. I don't understand the part
of performance statistically.



--
Sent from: https://weka.8497.n7.nabble.com/
_______________________________________________
Wekalist mailing list -- [hidden email]
Send posts to [hidden email]
To unsubscribe send an email to [hidden email]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/postorius/lists/wekalist.list.waikato.ac.nz
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html