Recurrent Neural Networks for Stochastic Control in Real-Time Bidding
This addresses the challenge of market uncertainty and penalties in real-time bidding for advertisers, representing an incremental improvement over existing methods.
The paper tackles the stochastic control problem in real-time bidding by proposing a Recurrent Neural Network (RNN) architecture that effectively provisions bids to avoid missing goals and deliberately falls short when costs exceed penalties.
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.