Pages: [1]
  Print  
Author Topic: GridParameterOptimization won't work with only one value  (Read 1263 times)
keith
Full Member
***
Posts: 160


« on: October 29, 2008, 08:40:19 PM »

I have a process where I am using GridParameterOptimization, where I intend to test a list of 5 or so values.  However, while I am developing the rest of the process, I want to have a single value in the list, to make sure the process runs successfully all the way through.  However, RapidMiner gives me an error when I try to run the process without multiple values to test:

Error in: GridParameterOptimization
No parameters were specified which should be optimized
You need to specify parameters of inner operators in order to allow the ParameterOptimization operation for optimization.

I realize this is a special case, as there is really no "optimizing" going on with only one value to test, but it seems like RM should execute the process with the single value, which obviously is chosen as "optimal".  Giving an error instead of a warning seems like an overreaction.  Am I missing something?

To be clear, this code fragment works:

        <operator name="GridParameterOptimization" class="GridParameterOptimization" expanded="yes">
            <list key="parameters">
              <parameter key="W-LWL.K"   value="1000,10,100"/>
            </list>

This one does not:

        <operator name="GridParameterOptimization" class="GridParameterOptimization" expanded="yes">
            <list key="parameters">
              <parameter key="W-LWL.K"   value="1000"/>
            </list>

Is this something fixable, or is there a reason I'm not aware of that explains this behavior?

Thanks,
Keith
Logged
Tobias Malbrecht
Global Moderator
Sr. Member
*****
Posts: 293



WWW
« Reply #1 on: October 30, 2008, 11:56:44 PM »

Hi Keith,

I realize this is a special case, as there is really no "optimizing" going on with only one value to test, but it seems like RM should execute the process with the single value, which obviously is chosen as "optimal".  Giving an error instead of a warning seems like an overreaction.  Am I missing something?

no, you are absolutely right. We should change that reaction in so far that RM only yields a warning message when only one parameter value is specified. Maybe we will be able to change that behaviour already for the next release which is about to come soon.

Regards,
Tobias
Logged

Tobias Malbrecht
Director of Product Marketing
RapidMiner
Pages: [1]
  Print  
 
Jump to: