Learning Piece-wise Linear Models from Large Scale Data for Ad Click Prediction
This provides an industrial-strength solution for ad click prediction, addressing scalability and efficiency in real-world business applications, though it appears incremental as it builds on existing piece-wise linear modeling approaches.
The paper tackles click-through rate prediction for online advertising by developing a piece-wise linear model that handles large-scale nonlinear sparse data, which has been deployed in Alibaba's system serving hundreds of millions of users daily since 2012.
CTR prediction in real-world business is a difficult machine learning problem with large scale nonlinear sparse data. In this paper, we introduce an industrial strength solution with model named Large Scale Piece-wise Linear Model (LS-PLM). We formulate the learning problem with $L_1$ and $L_{2,1}$ regularizers, leading to a non-convex and non-smooth optimization problem. Then, we propose a novel algorithm to solve it efficiently, based on directional derivatives and quasi-Newton method. In addition, we design a distributed system which can run on hundreds of machines parallel and provides us with the industrial scalability. LS-PLM model can capture nonlinear patterns from massive sparse data, saving us from heavy feature engineering jobs. Since 2012, LS-PLM has become the main CTR prediction model in Alibaba's online display advertising system, serving hundreds of millions users every day.