Menu

waffles_learn predict printf only labels predicted without input data

Help
maxim
2016-07-13
2016-07-17
  • maxim

    maxim - 2016-07-13

    hello there!

    it is said in usage information, that "waffles_learn predict" will print both features and predictions to stdout, but i only see one column in my arff and this is a prediction, with no features. am i doing something wrong ?

    λ waffles_learn predict model.json data_predict.arff
    @RELATION dataset

    @ATTRIBUTE my_feature real

    @DATA
    0.075
    0.04
    0
    0
    0
    0.015
    0

    data_predict.arff
    @RELATION dataset

    @ATTRIBUTE attr1 numeric
    @ATTRIBUTE attr2 numeric
    @ATTRIBUTE attr3 numeric
    @ATTRIBUTE attr4 numeric
    @ATTRIBUTE my_feature real

    @DATA
    0.918,0.233,0.491,0.648,0
    0.934,0.359,0.754,0.714,0
    0.252,0.605,0.148,0.146,0
    0.691,0.02,0.258,0.666,1
    0.037,0.897,0.304,0.146,0
    0.769,0.722,0.502,0.352,1
    0.222,0.172,0.425,0.488,0

     

    Last edit: maxim 2016-07-13
  • Mike Gashler

    Mike Gashler - 2016-07-13

    No, the documentation is wrong. I will update it.

    (If you want the features to be included, you can use
    waffles_transform mergehoriz dataset.arff data_predict.arff > both.arff
    to combine them into one file.)

     
  • maxim

    maxim - 2016-07-13

    Thank you very much, Mike!

     
  • maxim

    maxim - 2016-07-14

    tried using "waffles_transform mergehoriz" and was losing column type if it is "string". string columns are "real" in resulting dataset and contain a zero value.

    is there some workaround?

    file1.arff
    @RELATION file1

    @ATTRIBUTE id string

    @DATA
    "8560734"

    file2.arff
    @RELATION file2

    @ATTRIBUTE is_winner real

    @DATA
    0.285

    λ waffles_transform mergehoriz file1.arff file2.arff
    @RELATION Untitled

    @ATTRIBUTE id real
    @ATTRIBUTE is_winner real

    @DATA
    0,0.285

     
  • Mike Gashler

    Mike Gashler - 2016-07-14

    Unfortunately, I never finished adding support for string attributes to my GMatrix class, and I don't know of any work-around that would be simpler than adding this functionality to the GMatrix class. Sorry.

     
  • maxim

    maxim - 2016-07-15

    thanks for your answer, Mike!

    i made some changes, by converting string to integer, but still have issues. this time with "waffles_transform export". is this as it supposed to be - 8560734 changed to 8.56073e+006 ?

    i can live with scientific format, but it's obviously losing a digit:

    file3.arff
    @RELATION Untitled

    @ATTRIBUTE id numeric
    @ATTRIBUTE is_winner real

    @DATA
    8560734,0.285

    λ waffles_transform export file3.arff
    8.56073e+006,0.285

     
  • Mike Gashler

    Mike Gashler - 2016-07-16

    Thanks for pointing this out! It was using the default C++ stream precision, which is horrible. I changed it to use 14 digits by default and added a switch so you can specify the desired precision if you want more or less precision.

     
  • maxim

    maxim - 2016-07-17

    That's great, Mike! Thank you very much!

     

Anonymous
Anonymous

Add attachments
Cancel





Want the latest updates on software, tech news, and AI?
Get latest updates about software, tech news, and AI from SourceForge directly in your inbox once a month.