Du, M;
Cowen-Rivers, AI;
Wen, Y;
Sakulwongtana, P;
Wang, J;
Brorsson, M;
State, R;
(2020)
Know your enemies and know yourself in the real-time bidding function optimisation.
In:
Proceedings of the 2019 International Conference on Data Mining Workshops (ICDMW).
(pp. pp. 973-981).
IEEE: Beijing, China.
Preview |
Text
conference_041818.pdf - Accepted Version Download (592kB) | Preview |
Abstract
Real-time bidding (RTB) is a popular method to sell online ad space inventory using real-time auctions to determine which advertiser gets to make the ad impression. Advertisers can take user information into account when making their bids and get more control over the process. The goal of an optimal bidding function is to maximise the overall effectiveness of the ad campaigns defined by the advertisers under a certain budget constraint. A straightforward solution would be to model the bidding function in an explicit form. However, such functional solutions lack generality in practice and are insensitive to the stochastic behaviour of other bidders in the environment. In this paper, we propose to formulate the online auctions into a general mean field multi-agent framework, in which the agents compete with each other and each agent's best response strategy depends on its opponents' actions. We firstly introduce a novel Deep Attentive Survival Analysis (DASA) model to estimate the opponent's action distribution on the ad impression level which outperforms state-of-the-art survival analysis. Furthermore, we introduce the DASA model as the opponent model into the Mean Field Deep Deterministic Policy Gradients (DDPG) algorithm for each agent to learn the optimal bidding strategy and converge to the mean field equilibrium. The experiments have shown that with the inference of the market, the market converges to the equilibrium faster while playing against both fixed strategy agents and dynamic learning agents.
| Type: | Proceedings paper |
|---|---|
| Title: | Know your enemies and know yourself in the real-time bidding function optimisation |
| Event: | 2019 International Conference on Data Mining Workshops (ICDMW) |
| Open access status: | An open access version is available from UCL Discovery |
| DOI: | 10.1109/ICDMW.2019.00141 |
| Publisher version: | https://doi.org/10.1109/ICDMW.2019.00141 |
| Language: | English |
| Additional information: | This version is the author accepted manuscript. For information on re-use, please refer to the publisher’s terms and conditions. |
| Keywords: | Advertising data processing , electronic commerce , gradient methods , learning (artificial intelligence) , multi-agent systems , optimisation , stochastic processes , tendering |
| UCL classification: | UCL UCL > Provost and Vice Provost Offices > UCL BEAMS UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Engineering Science UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Engineering Science > Dept of Computer Science UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Engineering Science > UCL School of Management |
| URI: | https://discovery.ucl.ac.uk/id/eprint/10091626 |
Archive Staff Only
 |
View Item |
