"RapidMiner Tutorial Step 16 of 26"

kalexgann

The last paragraph of RapidMiner Tutorial Step 16 of 26 states:

Please note the MinMaxWrapper after the inner performance evaluator. This operator encapsulate the given performance criteria in such a way that no longer only average values but also minimum values are calculated during cross validation. Arbitrarily weighted linear combinations of the minimum and the normal average leads to far better generalization capabilities. Just change the weighting parameter to 0.0 or disable the operator in the context menu or delete it from the process to see the effect. The performance decreases rapidly when only the average performance is used as selection criterion.

When I disable/zero-out/remove the MinMaxWrapper ("FSMinMaxWrapper" (Performance(Min-Max))) the squared-error is reduced for the example set rather than increased as one would expect based on the tutorial text.  I adjusted the example set generator for more attributes/examples just to try a different scenario but still saw the squared-error improve by disabling the MinMaxWrapper, which is the opposite of what the tutorial suggests should happen.

Am I doing something wrong here?  I would expect that the tutorial examples would accurately reflect what happens when you follow them...please clarify.

Thanks!
Alex

land

Hi,
sorry for the inconvenience but this seems to be a problem in the text. I think it doesn't make much sense at all to use this min max wrapper...

Greetings,
Sebastian

kalexgann

Thanks for your reply, Sebastian!

Who developed the in-applcation tutorial/how would we request corrections?

land

Hi,
just send a bug report to our bug tracker. Someone will deal with it, if there's time.

Greetings,
  Sebastian