Building models with SAS Enterprise Miner, SAS Factory Miner, SAS Visual Data Mining and Machine Learning or just with programming

L2 regularization in SAS E-Miner

Accepted Solution Solved
Reply
Contributor
Posts: 43
Accepted Solution

L2 regularization in SAS E-Miner

Dear Mr,Mrs:

 

I want to apply l2 regularization for logistic regression but I found  only L1 regularization (LASSO). Can I get any technique or work around about this issue?.


Accepted Solutions
Solution
‎01-21-2016 09:59 PM
SAS Employee
Posts: 29

Re: L2 regularization in SAS E-Miner

[ Edited ]
Posted in reply to husseinmazaar

Hello,

 

As far as I know there is no proc that supports L2 regularization for logistic regression. PROC REG supports L2 regularization for linear regression (called RIDGE regression). One way to get around this is to treat your target as interval variable and use PROC REG; this regularizes the least squares loss function instead of the logistic loss function.  Below is the documentation of PROC REG that provides more information about the RIDGE option.

 

http://support.sas.com/documentation/cdl/en/statug/68162/HTML/default/viewer.htm#statug_reg_syntax08...

 

Hope this helps!

 

Funda

View solution in original post


All Replies
Solution
‎01-21-2016 09:59 PM
SAS Employee
Posts: 29

Re: L2 regularization in SAS E-Miner

[ Edited ]
Posted in reply to husseinmazaar

Hello,

 

As far as I know there is no proc that supports L2 regularization for logistic regression. PROC REG supports L2 regularization for linear regression (called RIDGE regression). One way to get around this is to treat your target as interval variable and use PROC REG; this regularizes the least squares loss function instead of the logistic loss function.  Below is the documentation of PROC REG that provides more information about the RIDGE option.

 

http://support.sas.com/documentation/cdl/en/statug/68162/HTML/default/viewer.htm#statug_reg_syntax08...

 

Hope this helps!

 

Funda

☑ This topic is solved.

Need further help from the community? Please ask a new question.

Discussion stats
  • 1 reply
  • 792 views
  • 0 likes
  • 2 in conversation