-
iMLCA: Machine Learning-powered Iterative Combinatorial Auctions with Interval Bidding
Authors:
Benjamin Lubin,
Sven Seuken,
Manuel Beyeler,
Gianluca Brero
Abstract:
Preference elicitation is a major challenge in large combinatorial auctions because the bundle space grows exponentially in the number of items. Recent work has used machine learning (ML) algorithms to identify a small set of bundles to query from each bidder. However, a shortcoming of this prior work is that bidders must submit exact values for the queried bundles, which can be quite costly. To a…
▽ More
Preference elicitation is a major challenge in large combinatorial auctions because the bundle space grows exponentially in the number of items. Recent work has used machine learning (ML) algorithms to identify a small set of bundles to query from each bidder. However, a shortcoming of this prior work is that bidders must submit exact values for the queried bundles, which can be quite costly. To address this, we propose iMLCA, a new ML-powered iterative combinatorial auction with interval bidding (i.e., where bidders submit upper and lower bounds instead of exact values). To steer the auction towards an efficient allocation, we introduce a price-based activity rule, asking bidders to tighten bounds on relevant bundles only. In our experiments, iMLCA achieves the same allocative efficiency as the prior ML-based auction that uses exact bidding. Moreover, it outperforms the well-known combinatorial clock auction in a realistically-sized domain.
△ Less
Submitted 29 August, 2021; v1 submitted 28 September, 2020;
originally announced September 2020.
-
Fourier Analysis-based Iterative Combinatorial Auctions
Authors:
Jakob Weissteiner,
Chris Wendler,
Sven Seuken,
Ben Lubin,
Markus Püschel
Abstract:
Recent advances in Fourier analysis have brought new tools to efficiently represent and learn set functions. In this paper, we bring the power of Fourier analysis to the design of combinatorial auctions (CAs). The key idea is to approximate bidders' value functions using Fourier-sparse set functions, which can be computed using a relatively small number of queries. Since this number is still too l…
▽ More
Recent advances in Fourier analysis have brought new tools to efficiently represent and learn set functions. In this paper, we bring the power of Fourier analysis to the design of combinatorial auctions (CAs). The key idea is to approximate bidders' value functions using Fourier-sparse set functions, which can be computed using a relatively small number of queries. Since this number is still too large for practical CAs, we propose a new hybrid design: we first use neural networks (NNs) to learn bidders' values and then apply Fourier analysis to the learned representations. On a technical level, we formulate a Fourier transform-based winner determination problem and derive its mixed integer program formulation. Based on this, we devise an iterative CA that asks Fourier-based queries. We experimentally show that our hybrid ICA achieves higher efficiency than prior auction designs, leads to a fairer distribution of social welfare, and significantly reduces runtime. With this paper, we are the first to leverage Fourier analysis in CA design and lay the foundation for future work in this area. Our code is available on GitHub: https://github.com/marketdesignresearch/FA-based-ICAs.
△ Less
Submitted 11 March, 2023; v1 submitted 22 September, 2020;
originally announced September 2020.
-
Machine Learning-powered Iterative Combinatorial Auctions
Authors:
Gianluca Brero,
Benjamin Lubin,
Sven Seuken
Abstract:
We present a machine learning-powered iterative combinatorial auction (MLCA). The main goal of integrating machine learning (ML) into the auction is to improve preference elicitation, which is a major challenge in large combinatorial auctions (CAs). In contrast to prior work, our auction design uses value queries instead of prices to drive the auction. The ML algorithm is used to help the auction…
▽ More
We present a machine learning-powered iterative combinatorial auction (MLCA). The main goal of integrating machine learning (ML) into the auction is to improve preference elicitation, which is a major challenge in large combinatorial auctions (CAs). In contrast to prior work, our auction design uses value queries instead of prices to drive the auction. The ML algorithm is used to help the auction decide which value queries to ask in every iteration. While using ML inside a CA introduces new challenges, we demonstrate how we obtain a design that is individually rational, satisfies no-deficit, has good incentives, and is computationally practical. We benchmark our new auction against the well-known combinatorial clock auction (CCA). Our results indicate that, especially in large domains, MLCA can achieve significantly higher allocative efficiency than the CCA, even with only a small number of value queries.
△ Less
Submitted 1 September, 2021; v1 submitted 18 November, 2019;
originally announced November 2019.
-
Computing Bayes-Nash Equilibria in Combinatorial Auctions with Verification
Authors:
Vitor Bosshard,
Benedikt Bünz,
Benjamin Lubin,
Sven Seuken
Abstract:
We present a new algorithm for computing pure-strategy $\varepsilon$-Bayes-Nash equilibria ($\varepsilon$-BNEs) in combinatorial auctions with continuous value and action spaces. An essential innovation of our algorithm is to separate the algorithm's search phase (for finding the $\varepsilon$-BNE) from the verification phase (for computing the $\varepsilon$). Using this approach, we obtain an alg…
▽ More
We present a new algorithm for computing pure-strategy $\varepsilon$-Bayes-Nash equilibria ($\varepsilon$-BNEs) in combinatorial auctions with continuous value and action spaces. An essential innovation of our algorithm is to separate the algorithm's search phase (for finding the $\varepsilon$-BNE) from the verification phase (for computing the $\varepsilon$). Using this approach, we obtain an algorithm that is both very fast and provides theoretical guarantees on the $\varepsilon$ it finds. Our main technical contribution is a verification method which allows us to upper bound the $\varepsilon$ across the whole continuous value space without making assumptions about the mechanism. Using our algorithm, we can now compute $\varepsilon$-BNEs in multi-minded domains that are significantly more complex than what was previously possible to solve. We release our code under an open-source license to enable researchers to perform algorithmic analyses of auctions, to enable bidders to analyze different strategies, and to facilitate many other applications.
△ Less
Submitted 24 July, 2020; v1 submitted 5 December, 2018;
originally announced December 2018.
-
A Bright Short Period M-M Eclipsing Binary from the KELT Survey: Magnetic Activity and the Mass-Radius Relationship for M-dwarfs
Authors:
Jack B. Lubin,
Joseph E. Rodriguez,
George Zhou,
Kyle E. Conroy,
Keivan G. Stassun,
Karen Collins,
Daniel J. Stevens,
Jonathan Labadie-Bartz,
Christopher Stockdale,
Gordon Myers,
Knicole D. Colón,
Joao Bento,
Petri Kehusmaa,
Romina Petrucci,
Emiliano Jofré,
Samuel N. Quinn,
Michael B. Lund,
Rudolf B. Kuhn,
Robert J. Siverd,
Thomas G. Beatty,
Caisey Harlingten,
Joshua Pepper,
B. Scott Gaudi,
David James,
Eric L. N. Jensen
, et al. (4 additional authors not shown)
Abstract:
We report the discovery of KELT J041621-620046, a moderately bright (J$\sim$10.2) M dwarf eclipsing binary system at a distance of 39$\pm$3 pc. KELT J041621-620046 was first identified as an eclipsing binary using observations from the Kilodegree Extremely Little Telescope (KELT) survey. The system has a short orbital period of $\sim$1.11 days and consists of components with M$_1$ =…
▽ More
We report the discovery of KELT J041621-620046, a moderately bright (J$\sim$10.2) M dwarf eclipsing binary system at a distance of 39$\pm$3 pc. KELT J041621-620046 was first identified as an eclipsing binary using observations from the Kilodegree Extremely Little Telescope (KELT) survey. The system has a short orbital period of $\sim$1.11 days and consists of components with M$_1$ = $0.447^{-0.047}_{+0.052}\,M_\odot$ and M$_2$ = $0.399^{-0.042}_{+0.046}\,M_\odot$ in nearly circular orbits. The radii of the two stars are R$_1$ = $0.540^{-0.032}_{+0.034}\,R_\odot$ and R$_2$ = $0.453\pm0.017\,R_\odot$. Full system and orbital properties were determined (to $\sim$10% error) by conducting an EBOP global modeling of the high precision photometric and spectroscopic observations obtained by the KELT Follow-up Network. Each star is larger by 17-28% and cooler by 4-10% than predicted by standard (non-magnetic) stellar models. Strong H$α$ emission indicates chromospheric activity in both stars. The observed radii and temperature discrepancies for both components are more consistent with those predicted by empirical relations that account for convective suppression due to magnetic activity.
△ Less
Submitted 7 June, 2017;
originally announced June 2017.
-
Games and Meta-Games: Pricing Rules for Combinatorial Mechanisms
Authors:
Benjamin Lubin
Abstract:
In settings where full incentive-compatibility is not available, such as core-constraint combinatorial auctions and budget-balanced combinatorial exchanges, we may wish to design mechanisms that are as incentive-compatible as possible. This paper offers a new characterization of approximate incentive-compatibility by casting the pricing problem as a meta-game between the center and the participati…
▽ More
In settings where full incentive-compatibility is not available, such as core-constraint combinatorial auctions and budget-balanced combinatorial exchanges, we may wish to design mechanisms that are as incentive-compatible as possible. This paper offers a new characterization of approximate incentive-compatibility by casting the pricing problem as a meta-game between the center and the participating agents. Through a suitable set of simplifications, we describe the equilibrium of this game as a variational problem. We use this to characterize the space of optimal prices, enabling closed-form solutions in restricted cases, and numerically-determined prices in the general case. We offer theory motivating this approach, and numerical experiments showing its application.
△ Less
Submitted 20 March, 2015;
originally announced March 2015.
-
Spectral goodness of fit for network models
Authors:
Jesse Shore,
Benjamin Lubin
Abstract:
We introduce a new statistic, 'spectral goodness of fit' (SGOF) to measure how well a network model explains the structure of an observed network. SGOF provides an absolute measure of fit, analogous to the standard R-squared in linear regression. Additionally, as it takes advantage of the properties of the spectrum of the graph Laplacian, it is suitable for comparing network models of diverse func…
▽ More
We introduce a new statistic, 'spectral goodness of fit' (SGOF) to measure how well a network model explains the structure of an observed network. SGOF provides an absolute measure of fit, analogous to the standard R-squared in linear regression. Additionally, as it takes advantage of the properties of the spectrum of the graph Laplacian, it is suitable for comparing network models of diverse functional forms, including both fitted statistical models and algorithmic generative models of networks. After introducing, defining, and providing guidance for interpreting SGOF, we illustrate the properties of the statistic with a number of examples and comparisons to existing techniques. We show that such a spectral approach to assessing model fit fills gaps left by earlier methods and can be widely applied.
△ Less
Submitted 27 July, 2014;
originally announced July 2014.
-
Communication Network Design: Balancing Modularity and Mixing via Optimal Graph Spectra
Authors:
Benjamin Lubin,
Jesse Shore,
Vatche Ishakian
Abstract:
By leveraging information technologies, organizations now have the ability to design their communication networks and crowdsourcing platforms to pursue various performance goals, but existing research on network design does not account for the specific features of social networks, such as the notion of teams. We fill this gap by demonstrating how desirable aspects of organizational structure can b…
▽ More
By leveraging information technologies, organizations now have the ability to design their communication networks and crowdsourcing platforms to pursue various performance goals, but existing research on network design does not account for the specific features of social networks, such as the notion of teams. We fill this gap by demonstrating how desirable aspects of organizational structure can be mapped parsimoniously onto the spectrum of the graph Laplacian allowing the specification of structural objectives and build on recent advances in non-convex programming to optimize them. This design framework is general, but we focus here on the problem of creating graphs that balance high modularity and low mixing time, and show how "liaisons" rather than brokers maximize this objective.
△ Less
Submitted 14 August, 2013;
originally announced August 2013.
-
Payment Rules through Discriminant-Based Classifiers
Authors:
Paul Duetting,
Felix Fischer,
Pitchayut Jirapinyo,
John K. Lai,
Benjamin Lubin,
David C. Parkes
Abstract:
In mechanism design it is typical to impose incentive compatibility and then derive an optimal mechanism subject to this constraint. By replacing the incentive compatibility requirement with the goal of minimizing expected ex post regret, we are able to adapt statistical machine learning techniques to the design of payment rules. This computational approach to mechanism design is applicable to dom…
▽ More
In mechanism design it is typical to impose incentive compatibility and then derive an optimal mechanism subject to this constraint. By replacing the incentive compatibility requirement with the goal of minimizing expected ex post regret, we are able to adapt statistical machine learning techniques to the design of payment rules. This computational approach to mechanism design is applicable to domains with multi-dimensional types and situations where computational efficiency is a concern. Specifically, given an outcome rule and access to a type distribution, we train a support vector machine with a special discriminant function structure such that it implicitly establishes a payment rule with desirable incentive properties. We discuss applications to a multi-minded combinatorial auction with a greedy winner-determination algorithm and to an assignment problem with egalitarian outcome rule. Experimental results demonstrate both that the construction produces payment rules with low ex post regret, and that penalizing classification errors is effective in preventing failures of ex post individual rationality.
△ Less
Submitted 6 August, 2012;
originally announced August 2012.
-
Quantifying the Strategyproofness of Mechanisms via Metrics on Payoff Distributions
Authors:
Benjamin Lubin,
David C. Parkes
Abstract:
Strategyproof mechanisms provide robust equilibrium with minimal assumptions about knowledge and rationality but can be unachievable in combination with other desirable properties such as budget-balance, stability against deviations by coalitions, and computational tractability. In the search for maximally-strategyproof mechanisms that simultaneously satisfy other desirable properties, we introduc…
▽ More
Strategyproof mechanisms provide robust equilibrium with minimal assumptions about knowledge and rationality but can be unachievable in combination with other desirable properties such as budget-balance, stability against deviations by coalitions, and computational tractability. In the search for maximally-strategyproof mechanisms that simultaneously satisfy other desirable properties, we introduce a new metric to quantify the strategyproofness of a mechanism, based on comparing the payoff distribution, given truthful reports, against that of a strategyproof "reference" mechanism that solves a problem relaxation. Focusing on combinatorial exchanges, we demonstrate that the metric is informative about the eventual equilibrium, where simple regretbased metrics are not, and can be used for online selection of an effective mechanism.
△ Less
Submitted 9 May, 2012;
originally announced May 2012.
