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.

We are pleased to invite you to contribute to the Eleventh ACM Conference on Recommender Systems (RecSys 2017), the premier venue for research and applications of recommendation technologies. The upcoming RecSys conference will be held in Como, Italy, from August, 27th to August 31st, 2017. The conference will continue RecSys' practice of connecting the research and practitioner communities to exchange ideas, frame problems, and share solutions. All the accepted papers will be published by ACM. We construe recommender systems broadly, including applications ranging from e-commerce to social networking, platforms from web to mobile and beyond, and a wide variety of technologies ranging from collaborative filtering to knowledge-based reasoning. The authors will be asked to indicate whether their paper is primarily Algorithm or Application focussed.