Rights statement: This is an Accepted Manuscript of an article published by Taylor & Francis in Journal of the Operational Research Society on 20 Feb 2019 available online: https://www.tandfonline.com/doi/full/10.1080/01605682.2018.1546650
Accepted author manuscript, 789 KB, PDF document
Available under license: CC BY-NC: Creative Commons Attribution-NonCommercial 4.0 International License
Final published version
Research output: Contribution to Journal/Magazine › Journal article › peer-review
Research output: Contribution to Journal/Magazine › Journal article › peer-review
}
TY - JOUR
T1 - Selecting Multiple Web Adverts - a Contextual Multi-armed Bandit with State Uncertainty
AU - Leslie, David Stuart
AU - Edwards, James Anthony
N1 - This is an Accepted Manuscript of an article published by Taylor & Francis in Journal of the Operational Research Society on 20 Feb 2019 available online: https://www.tandfonline.com/doi/full/10.1080/01605682.2018.1546650
PY - 2020/1/2
Y1 - 2020/1/2
N2 - We present a method to solve the problem of choosing a set of adverts to display to each of a sequence of web users. The objective is to maximise user clicks over time and to do so we must learn about the quality of each advert in an online manner by observing user clicks. We formulate the problem as a novel variant of a contextual combinatorial multi-armed bandit problem. The context takes the form of a probability distribution over the user's latent topic preference, and rewards are a particular nonlinear function of the selected set and the context. These features ensure that optimal sets of adverts are appropriately diverse. We give a flexible solution method which combines submodular optimisation with existing bandit index policies. User state uncertainty creates ambiguity in interpreting user feedback which prohibits exact Bayesian updating, but we give an approximate method that is shown to work well.
AB - We present a method to solve the problem of choosing a set of adverts to display to each of a sequence of web users. The objective is to maximise user clicks over time and to do so we must learn about the quality of each advert in an online manner by observing user clicks. We formulate the problem as a novel variant of a contextual combinatorial multi-armed bandit problem. The context takes the form of a probability distribution over the user's latent topic preference, and rewards are a particular nonlinear function of the selected set and the context. These features ensure that optimal sets of adverts are appropriately diverse. We give a flexible solution method which combines submodular optimisation with existing bandit index policies. User state uncertainty creates ambiguity in interpreting user feedback which prohibits exact Bayesian updating, but we give an approximate method that is shown to work well.
U2 - 10.1080/01605682.2018.1546650
DO - 10.1080/01605682.2018.1546650
M3 - Journal article
VL - 71
SP - 100
EP - 116
JO - Journal of the Operational Research Society
JF - Journal of the Operational Research Society
SN - 0160-5682
IS - 1
ER -