TY - GEN
T1 - Low-Regret Algorithms for Strategic Buyers with Unknown Valuations in Repeated Posted-Price Auctions
AU - Rhuggenaath, Jason
AU - Oliveira da Costa, Paulo Roberto de
AU - Zhang, Yingqian
AU - Akcay, Alp
AU - Kaymak, Uzay
PY - 2021
Y1 - 2021
N2 - We study repeated posted-price auctions where a single seller repeatedly interacts with a single buyer for a number of rounds. In previous works, it is common to consider that the buyer knows his own valuation with certainty. However, in many practical situations, the buyer may have a stochastic valuation. In this paper, we study repeated posted-price auctions from the perspective of a utility maximizing buyer who does not know the probability distribution of his valuation and only observes a sample from the valuation distribution after he purchases the item. We first consider non-strategic buyers and derive algorithms with sub-linear regret bounds that hold irrespective of the observed prices offered by the seller. These algorithms are then adapted into algorithms with similar guarantees for strategic buyers. We provide a theoretical analysis of our proposed algorithms and support our findings with numerical experiments. Our experiments show that, if the seller uses a low-regret algorithm for selecting the price, then strategic buyers can obtain much higher utilities compared to non-strategic buyers. Only when the prices of the seller are not related to the choices of the buyer, it is not beneficial to be strategic, but strategic buyers can still attain utilities of about 75% of the utility of non-strategic buyers.
AB - We study repeated posted-price auctions where a single seller repeatedly interacts with a single buyer for a number of rounds. In previous works, it is common to consider that the buyer knows his own valuation with certainty. However, in many practical situations, the buyer may have a stochastic valuation. In this paper, we study repeated posted-price auctions from the perspective of a utility maximizing buyer who does not know the probability distribution of his valuation and only observes a sample from the valuation distribution after he purchases the item. We first consider non-strategic buyers and derive algorithms with sub-linear regret bounds that hold irrespective of the observed prices offered by the seller. These algorithms are then adapted into algorithms with similar guarantees for strategic buyers. We provide a theoretical analysis of our proposed algorithms and support our findings with numerical experiments. Our experiments show that, if the seller uses a low-regret algorithm for selecting the price, then strategic buyers can obtain much higher utilities compared to non-strategic buyers. Only when the prices of the seller are not related to the choices of the buyer, it is not beneficial to be strategic, but strategic buyers can still attain utilities of about 75% of the utility of non-strategic buyers.
KW - No-regret learning
KW - Online learning
KW - Posted-price auctions
UR - http://www.scopus.com/inward/record.url?scp=85103303820&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-67661-2_25
DO - 10.1007/978-3-030-67661-2_25
M3 - Conference contribution
AN - SCOPUS:85103303820
SN - 9783030676605
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 416
EP - 436
BT - Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2020, Proceedings
A2 - Hutter, Frank
A2 - Kersting, Kristian
A2 - Lijffijt, Jefrey
A2 - Valera, Isabel
PB - Springer Science and Business Media Deutschland GmbH
T2 - European Conference on Machine Learning and Knowledge Discovery in Databases, ECML PKDD 2020
Y2 - 14 September 2020 through 18 September 2020
ER -