Web-Scale Bayesian Click-Through rate Prediction for Sponsored Search Advertising in Microsoft's Bing Search Engine

By Thore Graepel, Joaquin Quiñonero Candela, Thomas Borchert, Ralf Herbrich. 2010

619 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Counterfactual reasoning and learning systems: the example of computational advertising (2012, neutral): “The same analysis can be carried out for click prediction estimates expressed as arbitrary monotone combination of a position term γp(x) and an ad term βi(x), as shown, for instance, by Graepel et al. (2010).”
Practical Recommendations for Gradient-Based Training of Deep Architectures (2012, methodology): “A related formula turns up in Bayesian probit regression applied to sparse inputs (Graepel et al., 2010).”
A Tutorial on Thompson Sampling (2017, background): “Adaptations of Thompson sampling have now been successfully applied in a wide variety of domains, including revenue management [7], marketing [8], Monte Carlo tree search [9], A/B testing [10], Internet advertising [10, 11, 12], recommendation systems [13], hyperparameter tuning [14], and arcade gam”
Thompson Sampling for Contextual Bandits with Linear Payoffs (2012, methodology): “Several studies (e.g., Granmo (2010); Scott (2010); Graepel et al. (2010); Chapelle & Li (2011); May & Leslie (2011); Kaufmann et al. (2012)) have empirically demonstrated the eﬃcacy of TS: Scott (2010) provides a detailed discussion of probability matching techniques in many general settings along…”
Clipper: A Low-Latency Online Prediction Serving System (2016, background): “The past few years have seen an explosion of applications driven by machine learning, including recommendation systems [28, 60], voice assistants [18, 26, 55], and adtargeting [3,27].”
Concurrent Reinforcement Learning from Customer Interactions (2013, background): “In this setting, actions lead directly to immediate rewards, such as the click-through on an advert (Graepel et al., 2010), or news story (Li et al., 2010).”
Sequential Click Prediction for Sponsored Search with Recurrent Neural Networks (2014, methodology): “To measure the overall performance of each model, we follow the common practice in previous click prediction research in sponsored search and employ Area Under ROC Curve (AUC) and Relative Information Gain (RIG) as the evaluation metrics (Graepel et al. 2010; Xiong et al. 2012; Wang et al. 2013).”
Continual Learning with Bayesian Neural Networks for Non-Stationary Data (2020, methodology): “BF (Kulhavý & Zarrop, 1993) has been applied as an alternative to adaptation with an explicit transition model (e.g. Honkela & Valpola, 2003; Graepel et al., 2010).”
Automatic ad format selection via contextual bandits (2013, background): “A critical component of the auction is an estimate of the probability that an ad will be clicked on a given impression; also known as click-through rate (CTR) prediction [16, 1, 11, 6, 4, 15].”
Refining Recency Search Results with User Click Feedback (2011, methodology): “In particular, we have tried logistic regression and the probit-based regression [13], and observed similar performance as the linear model.”
Product-Based Neural Networks for User Response Prediction (2016, methodology): “Area under ROC Curve (AUC) and Relative Information Gain (RIG) are common evaluation metrics for response prediction accuracy a r X i v : 1611 .”
Convex Optimization: Algorithms and Complexity (2015, unknown): “…Observe outcome Yt and update Ppost 7: end for Recently, the TS algorithm has drawn considerable attention due to its state-of-the-art empirical performance [Scott, 2010, Chapelle and Li, 2011], which also led to its use in several industrial applications [Graepel et al., 2010, Tang et al., 2013].”
Simple Bayesian Algorithms for Best Arm Identification (2016, background): “In establishing these results, we exactly characterize the optimal rate of posterior convergence attainable by an adaptive algorithm, and provide interpretable bounds on this rate when measurement distributions are sub-Gaussian.”