Clustering for Private Interest-based Advertising

Epasto, Alessandro; Medina, Andrés Muñoz; Avery, S.; Bai, Yijian; Busa-Fekete, Róbert; Carey, CJ; Gao, Ya; Guthrie, David; Ghosh, Subham; Ioannidis, James; Jiao, Junyi; Łącki, Jakub; Lee, Jason; Mauser, Arne; Mirrokni, Vahab; Ravichandran, Deepak; Wei, Shi; Spero, Max; Sun, Yuxin; Syed, Umar; Vassilvitskii, Sergei; Wang, Shuo

doi:10.1145/3447548.3467180

Cited by 6 publications

(10 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In other words, each item is defined by d real-valued features and the seller reveals ℓ bits about the item to the buyer, as defined by a function that is unknown to the buyer. This model has been studied from an applied perspective in the context of ad auctions [Epasto et al, 2021]. We provide an algorithm with regret Õ((T dℓ) 1 /2 ).…”

Section: Our Resultsmentioning

confidence: 99%

“…Our theoretical model is motivated by recent work on designing privacyenhanced auctions for practical usage. An important application of these auctions is online advertising, where the items being auctioned are user queries and the auctioner must trade off between user privacy and revenue maximization [De Corniere and De Nijs, 2016, Rafieian and Yoganarasimhan, 2021, Epasto et al, 2021, Guha et al, 2011, Juels, 2001. Epasto et al [2021] present a detailed exploration of Chrome's Federated Learning of Cohorts (FLoC) API, where user information is masked.…”

Section: Our Resultsmentioning

confidence: 99%

“…An important application of these auctions is online advertising, where the items being auctioned are user queries and the auctioner must trade off between user privacy and revenue maximization [De Corniere and De Nijs, 2016, Rafieian and Yoganarasimhan, 2021, Epasto et al, 2021, Guha et al, 2011, Juels, 2001. Epasto et al [2021] present a detailed exploration of Chrome's Federated Learning of Cohorts (FLoC) API, where user information is masked. Our contribution is to provide a formal treatment of such auctions, including providing algorithms that have theoretical guarantees in the setting of arbitrary masking functions that are unknown to the buyer.…”

Section: Our Resultsmentioning

confidence: 99%

“…For instance, in online advertising auctions, bids represent how much advertisers are willing to pay to display their ad to a particular user. Historically, advertisers have bid based on uniquely identifying information about users, but there has been a growing effort to protect users' privacy by obfuscating this sensitive information [Juels, 2001, Guha et al, 2011, Epasto et al, 2021.…”

Section: Introductionmentioning

confidence: 99%

“…end if 10: end for Example 2.1 (Advertising auctions). We instantiate our model in the context of advertising auctions, taking inspiration from Epasto et al [2021]. The seller is a platform and the buyer is an advertiser.…”

mentioning

confidence: 99%

See 4 more Smart Citations

No-Regret Learning in Partially-Informed Auctions

Guo¹,

Jordan²,

Vitercik³

2022

Preprint

View full text Add to dashboard Cite

Auctions with partially-revealed information about items are broadly employed in real-world applications, but the underlying mechanisms have limited theoretical support. In this work, we study a machine learning formulation of these types of mechanisms, presenting algorithms that are no-regret from the buyer's perspective. Specifically, a buyer who wishes to maximize his utility interacts repeatedly with a platform over a series of T rounds. In each round, a new item is drawn from an unknown distribution and the platform publishes a price together with incomplete, "masked" information about the item. The buyer then decides whether to purchase the item. We formalize this problem as an online learning task where the goal is to have low regret with respect to a myopic oracle that has perfect knowledge of the distribution over items and the seller's masking function. When the distribution over items is known to the buyer and the mask is a SimHash function mapping R d to {0, 1} ℓ , our algorithm has regret Õ((T dℓ)1 /2 ). In a fully agnostic setting when the mask is an arbitrary function mapping to a set of size n, our algorithm has regret Õ(T 3 /4 n 1 /2 ). Finally, when the prices are stochastic, the algorithm has regret Õ((T n)1 /2 ).

show abstract

Section: Our Resultsmentioning

confidence: 99%

Section: Our Resultsmentioning

confidence: 99%

Section: Our Resultsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

mentioning

confidence: 99%

See 3 more Smart Citations

No-Regret Learning in Partially-Informed Auctions

Guo¹,

Jordan²,

Vitercik³

2022

Preprint

View full text Add to dashboard Cite

show abstract

Piloting personalization research through data-rich environments: a literature review and future research agenda

Mehmood

Verleye²,

Keyser³

et al. 2022

JOSM

View full text Add to dashboard Cite

PurposeOver the last 50 years, increased attention for personalization paved the way for one-to-one marketing efforts, but firms struggle to deliver on this promise. The purpose of this manuscript is to provide a complete picture on personalization, develop a future research agenda and put forth concrete advice on how to move the field forward from a theoretical, methodological, contextual, and practical viewpoint.Design/methodology/approachThis research follows a systematic literature review process, providing an in-depth analysis of 135 articles (covering 184 studies) to distill the (1) key building blocks and components of personalization and (2) theoretical, contextual, and methodological aspects of the studies.FindingsThis manuscript uncovers six personalization components that can be linked to two personalization building blocks: (1) learning: manner, transparency, and timing and (2) tailoring: touchpoints, level, and dynamics. For each of these components, the authors propose future research avenues to stimulate personalization research that accounts for challenges in today's data-rich environments (e.g. data privacy, dealing with new data types). A theoretical, contextual, and methodological (i.e. industry, country and personalization object) review of the selected studies leads to a set of concrete recommendations for future work: account for heterogeneity, embed theoretical perspectives, infuse methodological innovation, adopt appropriate evaluation metrics, and deal with legal/ethical challenges in data-rich environments. Finally, several managerial implications are put forth to support practitioners in their personalization efforts.Originality/valueThis research provides an integration of personalization research beyond existing and outdated review papers. Doing so, it accounts for the impact of new technologies and Artificial Intelligence and aims to advance the next generation of knowledge development on personalization.

show abstract

Feynman: Federated Learning-Based Advertising for Ecosystems-Oriented Mobile Apps Recommendation

Bian

Huang

et al. 2023

IEEE Trans. Serv. Comput.

View full text Add to dashboard Cite

Clustering for Private Interest-based Advertising

Cited by 6 publications

References 21 publications

No-Regret Learning in Partially-Informed Auctions

No-Regret Learning in Partially-Informed Auctions

Piloting personalization research through data-rich environments: a literature review and future research agenda

Feynman: Federated Learning-Based Advertising for Ecosystems-Oriented Mobile Apps Recommendation

Contact Info

Product

Resources

About