search

yinlou / mltk

Machine Learning Tool Kit
BSD 3-Clause "New" or "Revised" License
136 stars 74 forks source link

Fixing RandomForest.read() #10

Closed michaellavelle closed 8 years ago

michaellavelle commented 8 years ago

This PR fixes the following exception, occuring on RandomForest.read()

Exception in thread "main" java.lang.NumberFormatException: For input string: "mltk.predictor.tree.ensemble.rf.RandomForest]" at java.lang.NumberFormatException.forInputString(NumberFormatException.java:65) at java.lang.Integer.parseInt(Integer.java:492) at java.lang.Integer.parseInt(Integer.java:527) at mltk.predictor.tree.ensemble.rf.RandomForest.read(RandomForest.java:40)

yinlou commented 8 years ago

Hi @michaellavelle, this may not be a bug. In general, using PredictorReader and PredictorWriter is the recommended way to read and write models. I rely on reflection to determine the class of the model and therefore I need to read the line containing class information (such as, the first line), and when I've created the object, read() function will do the reset.

michaellavelle commented 8 years ago

Thanks for your reply @yinlou - I've now added Javadoc to the RandomForest.read() method to indicate that the method is intended to be used internally and that PredictorReader.read should be used instead, and I've removed the readLine() addition from RandomForest.read(). In the testing I've done, it seems the fix I added to the RandomForest.write() method for the line break is still needed - do let me know if there are any issues with this.

Thanks,

Michael

yinlou commented 8 years ago

Hi @michaellavelle, thanks for adding the comment. I think in stead of adding comments for each predictor, it's better to override the document for Predictor.read() and Predictor.write(). You can see the diff in latest version. To prevent confusion in the future, I will sketch out a wiki detailing how to work with this package. Thanks for you contribution!

michaellavelle commented 8 years ago

Thanks @yinlou - I've now updated our fork with your fixes - many thanks for all your work on this project