Due to recent updates, all users are required to create an Altair One account to login to the RapidMiner community. Click the Register button to create your account using the same email that you have previously used to login to the RapidMiner community. This will ensure that any previously created content will be synced to your Altair One account. Once you login, you will be asked to provide a username that identifies you to other Community users. Email us at Community with questions.

Multi-objective Optimization for Feature Selection

Asaber_07Asaber_07 Member Posts: 1 Learner I
in Help
thank you for the tutorial 

Multi-objective Optimization for Feature Selection

 I much appreciated, just a single question in the structural risk ie the number of features chosen what is its formula in terms of programming is what we do a ration the number chosen on the total number features ?? conceming the term C: trade off factor, its the interval between 0 and 1 ?? Thanks again

Answers

  • sgenzersgenzer Administrator, Moderator, Employee, RapidMiner Certified Analyst, Community Manager, Member, University Professor, PM Moderator Posts: 2,959 Community Manager
    cc'ing MOFS maestro @IngoRM
  • IngoRMIngoRM Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, Community Manager, RMResearcher, Member, University Professor Posts: 1,751 RM Founder
    Hi,
    just a single question in the structural risk ie the number of features chosen what is its formula in terms of programming is what we do a ration the number chosen on the total number features?
    While you could do this that way, this would only work well for feature selection, not for feature construction.  A good way to support both feature engineering tasks is simply to use the sum of features as the risk calculation.  This works well since the multi-objective optimizations does not care about the scale of the criterion at all and also does not need any trade-off parameters.

    conceming the term C: trade off factor, its the interval between 0 and 1 ?
    No, the range is unrestricted between 0 and infinite.  However, one of the beautiful things about a multi-objective regularization approach is that you actually do not need any C at all in the first place :-)

    Hope this helps,
    Ingo
Sign In or Register to comment.