WMRB: Learning to Rank in a Scalable Batch Training Approach
This work addresses scalability and accuracy issues in learning to rank for recommendation systems, representing an incremental improvement over existing methods.
The authors tackled the problem of improving learning to rank algorithms by proposing WMRB, which extends WARP with a new rank estimator and batch training, resulting in consistent performance gains and time efficiency in item recommendation tasks.
We propose a new learning to rank algorithm, named Weighted Margin-Rank Batch loss (WMRB), to extend the popular Weighted Approximate-Rank Pairwise loss (WARP). WMRB uses a new rank estimator and an efficient batch training algorithm. The approach allows more accurate item rank approximation and explicit utilization of parallel computation to accelerate training. In three item recommendation tasks, WMRB consistently outperforms WARP and other baselines. Moreover, WMRB shows clear time efficiency advantages as data scale increases.