Deep personalized re-targeting

Abstract

Predicting booking probability and value at the traveler level plays a central role in computational advertising for massive two-sided vacation rental marketplaces. These marketplaces host millions of travelers with long shopping cycles, spending a lot of time in the discovery phase. The footprint of the travelers in their discovery is a useful data source to help these marketplaces to predict shopping probability and value. However, there is no one-size-fits-all solution for this purpose. In this paper, we propose a hybrid model that infuses deep and shallow neural network embeddings into a gradient boosting tree model. This approach allows the latent preferences of millions of travelers to be automatically learned from sparse session logs. In addition, we present the architecture that we deployed into our production system. We find that there is a pragmatic sweet spot between expensive complex deep neural networks and simple shallow neural networks that can increase the prediction performance of a model by seven percent, based on offline analysis.

Wide Network Learning with Differential Privacy

“…This scenario is common in machine learning tasks. For example, in wide neural networks for language models and recommendation systems [MNHZB19,HMNZ19], the input data is usually extremely sparse, and only a small fraction of parameters are "active" in each update. Thus, the input sparsity usually produces a sparse gradient, which holds for a broad class of problems, as shown in Section 4.2.…”