Recurrent Neural Networks for Stochastic Control in Real-Time Bidding

TL;DR

This paper introduces an RNN-based approach for real-time bidding in online advertising, explicitly modeling market risk to improve campaign delivery under uncertain market conditions.

Contribution

It presents a novel RNN architecture that actively anticipates market shifts and balances delivery goals with cost penalties, advancing stochastic control in real-time bidding.

Findings

Effective in avoiding missed delivery goals

Balances cost and delivery penalties

Practical for production environments

Abstract

Bidding in real-time auctions can be a difficult stochastic control task; especially if underdelivery incurs strong penalties and the market is very uncertain. Most current works and implementations focus on optimally delivering a campaign given a reasonable forecast of the market. Practical implementations have a feedback loop to adjust and be robust to forecasting errors, but no implementation, to the best of our knowledge, uses a model of market risk and actively anticipates market shifts. Solving such stochastic control problems in practice is actually very challenging. This paper proposes an approximate solution based on a Recurrent Neural Network (RNN) architecture that is both effective and practical for implementation in a production environment. The RNN bidder provisions everything it needs to avoid missing its goal. It also deliberately falls short of its goal when buying the missing impressions would cost more than the penalty for not reaching it.