Multi-objective Ranking via Constrained Optimization
In this paper, we introduce an Augmented Lagrangian based method to incorporate the multiple objectives (MO) in a search ranking algorithm. Optimizing MOs is an essential and realistic requirement for building ranking models in production. The proposed method formulates MO in constrained optimization and solves the problem in the popular Boosting framework -- a novel contribution of our work. Furthermore, we propose a procedure to set up all optimization parameters in the problem. The experimental results show that the method successfully achieves MO criteria much more efficiently than existing methods.