Binary features in compressed instances

  • Nick

    Nick - 2012-03-07

    I am using the api to train a DecisionTree and then use it to classify.  I was having issues the tree was not working properly at all.  After doing some debugging and digging through the source I found that when an instance is compressed binary features always get assigned a weight of 1.  If I don't compress the instances then the tree trains and classifies just as expected.  Is this a bug in the compressed instances?  If not how are weights and binary features supposed to be used?


  • Frank Lin

    Frank Lin - 2012-03-09

    Hmm, maybe you've found a bug!
    Can you tell us where you believe the bug is located so we can quickly check to see what's wrong with it? Thanks :)


Log in to post a comment.