Volker

Hi Michael, that looks good so far, thank you for fixing it. The initial output shows now different metrics for test and train. I will let you know whether the test metric is actually computed for the test set in the end. Thank you very much for your help.

I was planning to go with the list-wise algorithms (AdaRank and LambdaMART).

Hi there, thank you for providing RankLib, this helps a lot to get into the practical aspects of Learning-to-Rank. It is great to have implementations of the most wiedely referenced methods available in the same package. I just have one problem: I can specify different metrics as metric2t (so for training) and metric2T (for testing), but it has no effect - training and testing get set to whatever I specify as metric2T. Why do I want to do that? I wanted to see if graded relevance actually helps my...

Hi there, thank you for providing RankLib, this helps a lot to get into the practical aspects of Learning-to-Rank. It is great to have implementations of the most wiedely referenced methods available in the same package. I just have one problem: I can specify different metrics as metric2t (so for training) and metric2T (for testing), but it has no effect - training and testing get set to whatever I specify as metric2T. Why do I want to do that? I wanted to see if graded relevance actually helps my...