Open
Description
Thanks for sharing the code. I have some small queries regarding your code to understand LambdaMart better.
First, You have three labels, i.e., 0, 1, 2. What are the meanings?
Second, why did you use decision tree regressors in LambdaMart? Should it not have binary classification instead?
Note that in the original paper, they suggested to improve it from RankNet, and the implementation was to solve whether a document pair is ranked correctly or not in terms of a query.
Metadata
Metadata
Assignees
Labels
No labels