Learning in Repeated Auctions

Nedelec, Thomas; Calauzènes, Clément; Karoui, Noureddine El; Perchet, Vianney

doi:10.1561/2200000077

Cited by 12 publications

(8 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…On the other hand, rival bidders can employ data-driven auto-bidding algorithms to optimize their own utilities. Moreover, the propensity of users to click an ad can vary over time due to exogenous factors, leading to incorrect utility estimations [19,34]. Observations in real-world data also support these conjectures, as detailed in the appendix.…”

Section: Methodsmentioning

confidence: 75%

“…At the heart of online advertising lies online auctions [17], where publishers repeatedly sell ad slots to advertisers seeking brand promotion, greater conversions, etc. Traditionally, incentive-compatible auctions such as second-price auctions are widely adopted, as they possess the desirable property of 'truthful bidding' for myopic bidders -truthfully revealing private values is optimal for these non-strategic bidders in order to maximize their immediate utility [32,34].…”

Section: Introductionmentioning

confidence: 99%

“…train and test conditions. This contradicts many real-world situations where the sellers and other rival bidders behave adversarially, as all parties seek to optimize their own utilities that are potentially conflicting with one another [34]. For instance, sellers may learn the distribution of bidders' private values and set personalized reserved prices in auctions [14,18,35].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Adversarial Constrained Bidding via Minimax Regret Optimization with Causality-Aware Reinforcement Learning

Wang

Fang

et al. 2023

Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining

View full text Add to dashboard Cite

The proliferation of the Internet has led to the emergence of online advertising, driven by the mechanics of online auctions. In these repeated auctions, software agents participate on behalf of aggregated advertisers to optimize for their long-term utility. To fulfill the diverse demands, bidding strategies are employed to optimize advertising objectives subject to different spending constraints. Existing approaches on constrained bidding typically rely on i.i.d. train and test conditions, which contradicts the adversarial nature of online ad markets where different parties possess potentially conflicting objectives. In this regard, we explore the problem of constrained bidding in adversarial bidding environments, which assumes no knowledge about the adversarial factors. Instead of relying on the i.i.d. assumption, our insight is to align the train distribution of environments with the potential test distribution meanwhile minimizing policy regret. Based on this insight, we propose a practical Minimax Regret Optimization (MiRO) approach that interleaves between a teacher finding adversarial environments for tutoring and a learner meta-learning its policy over the given distribution of environments. In addition, we pioneer to incorporate expert demonstrations for learning bidding strategies. Through a causality-aware policy design, we improve upon MiRO by distilling knowledge from the experts. Extensive experiments on both industrial data and synthetic data show that our method, MiRO with Causality-aware reinforcement Learning (MiROCL), outperforms prior methods by over 30%.

show abstract

Section: Methodsmentioning

confidence: 75%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Adversarial Constrained Bidding via Minimax Regret Optimization with Causality-Aware Reinforcement Learning

Wang

Fang

et al. 2023

Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining

View full text Add to dashboard Cite

show abstract

“…These examples are all drawn from public policy, where there is an intrinsic concern for social welfare. This contrasts with commercial applications, where the goal is typically to maximize (directly observable) profits by monopolist pricing (den Boer, 2015), or more generally by reserve price setting in auctions (Nedelec et al, 2022). Adaptive pricing algorithms are used in applications such as online ad auctions.…”

Section: Discussionmentioning

confidence: 99%

Adaptive Maximization of Social Welfare

Cesa-Bianchi,

Colomboni,

Kasy

2024

Preprint

View full text Add to dashboard Cite

We consider the problem of repeatedly choosing policies to maximize social welfare. Welfare is a weighted sum of private utility and public revenue. Earlier outcomes inform later policies. Utility is not observed, but indirectly inferred. Response functions are learned through experimentation. We derive a lower bound on regret, and a matching adversarial upper bound for a variant of the Exp3 algorithm. Cumulative regret grows at a rate of T2/3. This implies that (i) welfare maximization is harder than the multi-armed bandit problem (with a rate of T1/2 for finite policy sets), and (ii) our algorithm achieves the optimal rate. For the stochastic setting, if social welfare is concave, we can achieve a rate of T1/2 (for continuous policy sets), using a dyadic search algorithm. We analyze an extension to nonlinear income taxation, and sketch an extension to commodity taxation. We compare our setting to monopoly pricing (which is easier), and price setting for bilateral trade (which is harder). (Stone Center on Socio-Economic Inequality Working Paper)

show abstract

“…T HE artificial intelligence (AI) algorithms have shown to be successful in a variety of fields and jobs, from tiny molecules [1], to identifying exoplanets [2] going through commerce [3]. The popularity of machine learning (ML) has been boosted as a result of the success of deep learning (DL) [4]- [6].…”

Section: Plain I Introductionmentioning

confidence: 99%

Mathematical Modeling and Optimal Stopping Theory-based additional layers for 30-Day Rate Risk Prediction of Readmission to Intensive Care Units

hassouny¹,

Bouanani²,

Qaraqe³

2022

Preprint

View full text Add to dashboard Cite

The importance of 30-day patients’ readmissions (PRs) to intensive unit care stems from the significant cost and mortality risk when the patient’s chosen class (i.e., readmitted or not to the hospital) is incorrect. The overall accuracy (OA) of the PRs classification obtained in the literature is still moderate, particularly for machine learning (ML)-enabled ANNs, where OA is around 65%, resulting in 35% critical wrong decisions. To improve such an OA, a three-stage ML-assisted algorithm employing both support vector machines (SVMs) and artificial neural networks (ANNs) techniques is proposed. Starting with a well-fitted PR accuracies distribution and using mathematical modeling, the global optimal accuracies’ interval of misclassified patients (OIMP) is obtained, and a novel theorem for the generalized secretary problem is introduced and used to select the most likely well-classified patients. The algorithm’s final phase consists of flipping the remaining incorrectly classified patients in the OIMP. Using the presented approach with ANN and SVM, the OA was raised by 5% and 19%, respectively. The proposed approach may be used for any binary classification application and expanded to any multi-class problem.

show abstract

Learning in Repeated Auctions

Cited by 12 publications

References 0 publications

Adversarial Constrained Bidding via Minimax Regret Optimization with Causality-Aware Reinforcement Learning

Adversarial Constrained Bidding via Minimax Regret Optimization with Causality-Aware Reinforcement Learning

Adaptive Maximization of Social Welfare

Mathematical Modeling and Optimal Stopping Theory-based additional layers for 30-Day Rate Risk Prediction of Readmission to Intensive Care Units

Contact Info

Product

Resources

About