#13 Fixup Tokenizer Trainer to include cutoff and iterations

closed-accepted
nobody
None
5
2010-07-14
2010-07-14
James Kosin
No

This patch adds cutoff and iteration parameters to the training of the model.

I'm still trying to determine if we should pass in the parameters class into the train() function call or keep passing the individual parameters to all the classes for training. Either way, I still have a lot of work to migrate the generic parameters to all the training sets....

Discussion

  • James Kosin
    James Kosin
    2010-07-14

    Tokenizer Trainer Patch (take 1)

     
  • Joern Kottmann
    Joern Kottmann
    2010-07-14

    Thanks, nice patch. Additionally I also made cutoff and iterations configurable for the cross validator. I still think many people do not really want to adjust these parameters, and maybe are not aware that we always use 5 and 100. For this reason its nice to have an overloaded train method which uses these as default.

    When you add parameters please also add them to the javadoc comment.

    "I'm still trying to determine if we should pass in the parameters class into the train() function call or keep passing the individual parameters to all the classes for training. Either way, I still have a lot of work to migrate the generic parameters to all the training sets...."

    Not sure what you mean here exactly. Right now we are just passing the parameters to the train method of the different components, which other way to provide these parameters, do you have in mind ?

    Thanks again for the patch,
    Jörn

     
  • Joern Kottmann
    Joern Kottmann
    2010-07-14

    • status: open --> closed-accepted