libFM output

[masha....@gmail.com]

Hi! I'm new to LibFM. Got the following output and concerned with huge discrepancy between test and train:

Final\tTrain=0.80606\tTest=0.370107

roc_auc_score is 0.82932847197120885

This is my libFM call:
./libFM -task c -method sgd -save_model model1 -learn_rate 0.01 -regular '0,0,0.01' -init_stdev 0.1 -train $train_file -test $test_file -iter 500 -dim '1,1,10' -out output.libfm

Is my concern valid and if yes, how do I go about fixing that gap?

Thanks!

[Thierry Silbermann]

It might be a bad case of overfitting.
How many features do you have for this problem?
What is the number of samples in train and test ?
Does train score decrease during training ? What about test ?

You might want to try increasing the -regular parameters.