Learning Revenue-Maximizing Auctions With Differentiable Matching

Michael J. Curry, Uro Lyi, Tom Goldstein, John P. Dickerson
Proceedings of The 25th International Conference on Artificial Intelligence and Statistics, PMLR 151:6062-6073, 2022.

Abstract

We propose a new architecture to approximately learn incentive compatible, revenue-maximizing auctions from sampled valuations. Our architecture uses the Sinkhorn algorithm to perform a differentiable bipartite matching which allows the network to learn strategyproof revenue-maximizing mechanisms in settings not learnable by the previous RegretNet architecture. In particular, our architecture is able to learn mechanisms in settings without free disposal where each bidder must be allocated exactly some number of items. In experiments, we show our approach successfully recovers multiple known optimal mechanisms and high-revenue, low-regret mechanisms in larger settings where the optimal mechanism is unknown.

Cite this Paper

BibTeX
@InProceedings{pmlr-v151-curry22a, title = { Learning Revenue-Maximizing Auctions With Differentiable Matching }, author = {Curry, Michael J. and Lyi, Uro and Goldstein, Tom and Dickerson, John P.}, booktitle = {Proceedings of The 25th International Conference on Artificial Intelligence and Statistics}, pages = {6062--6073}, year = {2022}, editor = {Camps-Valls, Gustau and Ruiz, Francisco J. R. and Valera, Isabel}, volume = {151}, series = {Proceedings of Machine Learning Research}, month = {28--30 Mar}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v151/curry22a/curry22a.pdf}, url = {https://proceedings.mlr.press/v151/curry22a.html}, abstract = { We propose a new architecture to approximately learn incentive compatible, revenue-maximizing auctions from sampled valuations. Our architecture uses the Sinkhorn algorithm to perform a differentiable bipartite matching which allows the network to learn strategyproof revenue-maximizing mechanisms in settings not learnable by the previous RegretNet architecture. In particular, our architecture is able to learn mechanisms in settings without free disposal where each bidder must be allocated exactly some number of items. In experiments, we show our approach successfully recovers multiple known optimal mechanisms and high-revenue, low-regret mechanisms in larger settings where the optimal mechanism is unknown. } }
Endnote
%0 Conference Paper %T Learning Revenue-Maximizing Auctions With Differentiable Matching %A Michael J. Curry %A Uro Lyi %A Tom Goldstein %A John P. Dickerson %B Proceedings of The 25th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2022 %E Gustau Camps-Valls %E Francisco J. R. Ruiz %E Isabel Valera %F pmlr-v151-curry22a %I PMLR %P 6062--6073 %U https://proceedings.mlr.press/v151/curry22a.html %V 151 %X We propose a new architecture to approximately learn incentive compatible, revenue-maximizing auctions from sampled valuations. Our architecture uses the Sinkhorn algorithm to perform a differentiable bipartite matching which allows the network to learn strategyproof revenue-maximizing mechanisms in settings not learnable by the previous RegretNet architecture. In particular, our architecture is able to learn mechanisms in settings without free disposal where each bidder must be allocated exactly some number of items. In experiments, we show our approach successfully recovers multiple known optimal mechanisms and high-revenue, low-regret mechanisms in larger settings where the optimal mechanism is unknown.
APA
Curry, M.J., Lyi, U., Goldstein, T. & Dickerson, J.P.. (2022). Learning Revenue-Maximizing Auctions With Differentiable Matching . Proceedings of The 25th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 151:6062-6073 Available from https://proceedings.mlr.press/v151/curry22a.html.

Related Material