A Context-Integrated Transformer-Based Neural Network for Auction Design

Zhijian Duan, Jingwu Tang, Yutong Yin, Zhe Feng, Xiang Yan, Manzil Zaheer, Xiaotie Deng
Proceedings of the 39th International Conference on Machine Learning, PMLR 162:5609-5626, 2022.

Abstract

One of the central problems in auction design is developing an incentive-compatible mechanism that maximizes the auctioneer’s expected revenue. While theoretical approaches have encountered bottlenecks in multi-item auctions, recently, there has been much progress on finding the optimal mechanism through deep learning. However, these works either focus on a fixed set of bidders and items, or restrict the auction to be symmetric. In this work, we overcome such limitations by factoring public contextual information of bidders and items into the auction learning framework. We propose $\mathtt{CITransNet}$, a context-integrated transformer-based neural network for optimal auction design, which maintains permutation-equivariance over bids and contexts while being able to find asymmetric solutions. We show by extensive experiments that $\mathtt{CITransNet}$ can recover the known optimal solutions in single-item settings, outperform strong baselines in multi-item auctions, and generalize well to cases other than those in training.

Cite this Paper

BibTeX
@InProceedings{pmlr-v162-duan22a, title = {A Context-Integrated Transformer-Based Neural Network for Auction Design}, author = {Duan, Zhijian and Tang, Jingwu and Yin, Yutong and Feng, Zhe and Yan, Xiang and Zaheer, Manzil and Deng, Xiaotie}, booktitle = {Proceedings of the 39th International Conference on Machine Learning}, pages = {5609--5626}, year = {2022}, editor = {Chaudhuri, Kamalika and Jegelka, Stefanie and Song, Le and Szepesvari, Csaba and Niu, Gang and Sabato, Sivan}, volume = {162}, series = {Proceedings of Machine Learning Research}, month = {17--23 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v162/duan22a/duan22a.pdf}, url = {https://proceedings.mlr.press/v162/duan22a.html}, abstract = {One of the central problems in auction design is developing an incentive-compatible mechanism that maximizes the auctioneer’s expected revenue. While theoretical approaches have encountered bottlenecks in multi-item auctions, recently, there has been much progress on finding the optimal mechanism through deep learning. However, these works either focus on a fixed set of bidders and items, or restrict the auction to be symmetric. In this work, we overcome such limitations by factoring public contextual information of bidders and items into the auction learning framework. We propose $\mathtt{CITransNet}$, a context-integrated transformer-based neural network for optimal auction design, which maintains permutation-equivariance over bids and contexts while being able to find asymmetric solutions. We show by extensive experiments that $\mathtt{CITransNet}$ can recover the known optimal solutions in single-item settings, outperform strong baselines in multi-item auctions, and generalize well to cases other than those in training.} }
Endnote
%0 Conference Paper %T A Context-Integrated Transformer-Based Neural Network for Auction Design %A Zhijian Duan %A Jingwu Tang %A Yutong Yin %A Zhe Feng %A Xiang Yan %A Manzil Zaheer %A Xiaotie Deng %B Proceedings of the 39th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2022 %E Kamalika Chaudhuri %E Stefanie Jegelka %E Le Song %E Csaba Szepesvari %E Gang Niu %E Sivan Sabato %F pmlr-v162-duan22a %I PMLR %P 5609--5626 %U https://proceedings.mlr.press/v162/duan22a.html %V 162 %X One of the central problems in auction design is developing an incentive-compatible mechanism that maximizes the auctioneer’s expected revenue. While theoretical approaches have encountered bottlenecks in multi-item auctions, recently, there has been much progress on finding the optimal mechanism through deep learning. However, these works either focus on a fixed set of bidders and items, or restrict the auction to be symmetric. In this work, we overcome such limitations by factoring public contextual information of bidders and items into the auction learning framework. We propose $\mathtt{CITransNet}$, a context-integrated transformer-based neural network for optimal auction design, which maintains permutation-equivariance over bids and contexts while being able to find asymmetric solutions. We show by extensive experiments that $\mathtt{CITransNet}$ can recover the known optimal solutions in single-item settings, outperform strong baselines in multi-item auctions, and generalize well to cases other than those in training.
APA
Duan, Z., Tang, J., Yin, Y., Feng, Z., Yan, X., Zaheer, M. & Deng, X.. (2022). A Context-Integrated Transformer-Based Neural Network for Auction Design. Proceedings of the 39th International Conference on Machine Learning, in Proceedings of Machine Learning Research 162:5609-5626 Available from https://proceedings.mlr.press/v162/duan22a.html.

Related Material