Waterfall ranking and filtering optimization  
Author Liang Dai


Co-Author(s) Ram Akella


Abstract In this paper, we discuss about how to efficiently select a subset of items from a large candidate pool through waterfall filtering system. Firstly, we discuss about why waterfall filtering systems are widely used in industry. Then, we dive into the problem of how to find the optimal settings of the system given Gaussian inputs with unbiased predictors. To solve that problem, we start from the analysis of only one single waterfall ranking and filtering problem, and then extend our analysis to multiple waterfalls case. After that, we briefly discuss about the how to take infra and latency cost into consideration. In the simulation, our proposed method shows to be accurate and efficient in choosing the optimal settings of a waterfall ranking and filtering system.


Keywords Waterfall filtering, Large scale selection system
    Article #:  DSIS19-100
Proceedings of ISSAT International Conference on Data Science & Intelligent Systems
August 1-3, 2019 - Las Vegas, NV, U.S.A.