arXiv subjects

Amy Greenwald

Publications and source records attributed to Amy Greenwald.

At least 37 records · Page 2Linked to original sources

arXiv2022

Regret Pruning for Learning Equilibria in Simulation-Based Games

In recent years, empirical game-theoretic analysis (EGTA) has emerged as a powerful tool for analyzing games in which an exact specification of the utilities is unavailable. Instead, EGTA assumes access to an oracle, i.e., a simulator, which can generate unbiased noisy samples of players' unknown utilities, given a strategy profile. Utilities can thus be empirically estimated by repeatedly querying the simulator. Recently, various progressive sampling (PS) algorithms have been proposed, which aim to produce PAC-style learning guarantees (e.g., approximate Nash equilibria with high probability) using as few simulator queries as possible. One recent work introduces a pruning technique called regret-pruning which further minimizes the number of simulator queries placed in PS algorithms which aim to learn pure Nash equilibria. In this paper, we address a serious limitation of this original regret pruning approach -- it is only able to guarantee that true pure Nash equilibria of the empirical game are approximate equilibria of the true game, and is unable to provide any strong guarantees regarding the efficacy of approximate pure Nash equilibria. This is a significant limitation since in many games, pure Nash equilibria are computationally intractable to find, or even non-existent. We introduce three novel regret pruning variations. The first two variations generalize the original regret pruning approach to yield guarantees for approximate pure Nash equilibria of the empirical game. The third variation goes further to even yield strong guarantees for all approximate mixed Nash equilibria of the empirical game. We use these regret pruning variations to design two novel progressive sampling algorithms, PS-REG+ and PS-REG-M, which experimentally outperform the previous state-of-the-art algorithms for learning pure and mixed equilibria, respectively, of simulation-based games.

cs.GT↗

arXiv2022

Zero-Sum Stochastic Stackelberg Games

Zero-sum stochastic games have found important applications in a variety of fields, from machine learning to economics. Work on this model has primarily focused on the computation of Nash equilibrium due to its effectiveness in solving adversarial board and video games. Unfortunately, a Nash equilibrium is not guaranteed to exist in zero-sum stochastic games when the payoffs at each state are not convex-concave in the players' actions. A Stackelberg equilibrium, however, is guaranteed to exist. Consequently, in this paper, we study zero-sum stochastic Stackelberg games. Going beyond known existence results for (non-stationary) Stackelberg equilibria, we prove the existence of recursive (i.e., Markov perfect) Stackelberg equilibria (recSE) in these games, provide necessary and sufficient conditions for a policy profile to be a recSE, and show that recSE can be computed in (weakly) polynomial time via value iteration. Finally, we show that zero-sum stochastic Stackelberg games can model the problem of pricing and allocating goods across agents and time. More specifically, we propose a zero-sum stochastic Stackelberg game whose recSE correspond to the recursive competitive equilibria of a large class of stochastic Fisher markets. We close with a series of experiments that showcase how our methodology can be used to solve the consumption-savings problem in stochastic Fisher markets.

cs.GT↗

arXiv2022

Exploitability Minimization in Games and Beyond

Pseudo-games are a natural and well-known generalization of normal-form games, in which the actions taken by each player affect not only the other players' payoffs, as in games, but also the other players' strategy sets. The solution concept par excellence for pseudo-games is the generalized Nash equilibrium (GNE), i.e., a strategy profile at which each player's strategy is feasible and no player can improve their payoffs by unilaterally deviating to another strategy in the strategy set determined by the other players' strategies. The computation of GNE in pseudo-games has long been a problem of interest, due to applications in a wide variety of fields, from environmental protection to logistics to telecommunications. Although computing GNE is PPAD-hard in general, it is still of interest to try to compute them in restricted classes of pseudo-games. One approach is to search for a strategy profile that minimizes exploitability, i.e., the sum of the regrets across all players. As exploitability is nondifferentiable in general, developing efficient first-order methods that minimize it might not seem possible at first glance. We observe, however, that the exploitability-minimization problem can be recast as a min-max optimization problem, and thereby obtain polynomial-time first-order methods to compute a refinement of GNE, namely the variational equilibria (VE), in convex-concave cumulative regret pseudo-games with jointly convex constraints. More generally, we also show that our methods find the stationary points of the exploitability in polynomial time in Lipschitz-smooth pseudo-games with jointly convex constraints. Finally, we demonstrate in experiments that our methods not only outperform known algorithms, but that even in pseudo-games where they are not guaranteed to converge to a GNE, they may do so nonetheless, with proper initialization.

cs.GT↗

arXiv2022

Gradient Descent Ascent in Min-Max Stackelberg Games

Min-max optimization problems (i.e., min-max games) have attracted a great deal of attention recently as their applicability to a wide range of machine learning problems has become evident. In this paper, we study min-max games with dependent strategy sets, where the strategy of the first player constrains the behavior of the second. Such games are best understood as sequential, i.e., Stackelberg, games, for which the relevant solution concept is Stackelberg equilibrium, a generalization of Nash. One of the most popular algorithms for solving min-max games is gradient descent ascent (GDA). We present a straightforward generalization of GDA to min-max Stackelberg games with dependent strategy sets, but show that it may not converge to a Stackelberg equilibrium. We then introduce two variants of GDA, which assume access to a solution oracle for the optimal Karush Kuhn Tucker (KKT) multipliers of the games' constraints. We show that such an oracle exists for a large class of convex-concave min-max Stackelberg games, and provide proof that our GDA variants with such an oracle converge in $O(\frac{1}{\varepsilon^2})$ iterations to an $\varepsilon$-Stackelberg equilibrium, improving on the most efficient algorithms currently known which converge in $O(\frac{1}{\varepsilon^3})$ iterations. We then show that solving Fisher markets, a canonical example of a min-max Stackelberg game, using our novel algorithm, corresponds to buyers and sellers using myopic best-response dynamics in a repeated market, allowing us to prove the convergence of these dynamics in $O(\frac{1}{\varepsilon^2})$ iterations in Fisher markets. We close by describing experiments on Fisher markets which suggest potential ways to extend our theoretical results, by demonstrating how different properties of the objective function can affect the convergence and convergence rate of our algorithms.

cs.GT↗

arXiv2022

Computational and Data Requirements for Learning Generic Properties of Simulation-Based Games

Empirical game-theoretic analysis (EGTA) is primarily focused on learning the equilibria of simulation-based games. Recent approaches have tackled this problem by learning a uniform approximation of the game's utilities, and then applying precision-recall theorems: i.e., all equilibria of the true game are approximate equilibria in the estimated game, and vice-versa. In this work, we generalize this approach to all game properties that are well behaved (i.e., Lipschitz continuous in utilities), including regret (which defines Nash and correlated equilibria), adversarial values, and power-mean and Gini social welfare. Further, we introduce a novel algorithm -- progressive sampling with pruning (PSP) -- for learning a uniform approximation and thus any well-behaved property of a game, which prunes strategy profiles once the corresponding players' utilities are well-estimated, and we analyze its data and query complexities in terms of the a priori unknown utility variances. We experiment with our algorithm extensively, showing that 1) the number of queries that PSP saves is highly sensitive to the utility variance distribution, and 2) PSP consistently outperforms theoretical upper bounds, achieving significantly lower query complexities than natural baselines. We conclude with experiments that uncover some of the remaining difficulties with learning properties of simulation-based games, in spite of recent advances in statistical EGTA methodology, including those developed herein.

cs.GT↗

arXiv2022

Hindsight and Sequential Rationality of Correlated Play

Driven by recent successes in two-player, zero-sum game solving and playing, artificial intelligence work on games has increasingly focused on algorithms that produce equilibrium-based strategies. However, this approach has been less effective at producing competent players in general-sum games or those with more than two players than in two-player, zero-sum games. An appealing alternative is to consider adaptive algorithms that ensure strong performance in hindsight relative to what could have been achieved with modified behavior. This approach also leads to a game-theoretic analysis, but in the correlated play that arises from joint learning dynamics rather than factored agent behavior at equilibrium. We develop and advocate for this hindsight rationality framing of learning in general sequential decision-making settings. To this end, we re-examine mediated equilibrium and deviation types in extensive-form games, thereby gaining a more complete understanding and resolving past misconceptions. We present a set of examples illustrating the distinct strengths and weaknesses of each type of equilibrium in the literature, and prove that no tractable concept subsumes all others. This line of inquiry culminates in the definition of the deviation and equilibrium classes that correspond to algorithms in the counterfactual regret minimization (CFR) family, relating them to all others in the literature. Examining CFR in greater detail further leads to a new recursive definition of rationality in correlated play that extends sequential rationality in a way that naturally applies to hindsight evaluation.

cs.GT↗

arXiv2022

Efficient Deviation Types and Learning for Hindsight Rationality in Extensive-Form Games

Hindsight rationality is an approach to playing general-sum games that prescribes no-regret learning dynamics for individual agents with respect to a set of deviations, and further describes jointly rational behavior among multiple agents with mediated equilibria. To develop hindsight rational learning in sequential decision-making settings, we formalize behavioral deviations as a general class of deviations that respect the structure of extensive-form games. Integrating the idea of time selection into counterfactual regret minimization (CFR), we introduce the extensive-form regret minimization (EFR) algorithm that achieves hindsight rationality for any given set of behavioral deviations with computation that scales closely with the complexity of the set. We identify behavioral deviation subsets, the partial sequence deviation types, that subsume previously studied types and lead to efficient EFR instances in games with moderate lengths. In addition, we present a thorough empirical analysis of EFR instantiated with different deviation types in benchmark games, where we find that stronger types typically induce better performance.

cs.GT↗

arXiv2022

Interpolating Between Softmax Policy Gradient and Neural Replicator Dynamics with Capped Implicit Exploration

Neural replicator dynamics (NeuRD) is an alternative to the foundational softmax policy gradient (SPG) algorithm motivated by online learning and evolutionary game theory. The NeuRD expected update is designed to be nearly identical to that of SPG, however, we show that the Monte Carlo updates differ in a substantial way: the importance correction accounting for a sampled action is nullified in the SPG update, but not in the NeuRD update. Naturally, this causes the NeuRD update to have higher variance than its SPG counterpart. Building on implicit exploration algorithms in the adversarial bandit setting, we introduce capped implicit exploration (CIX) estimates that allow us to construct NeuRD-CIX, which interpolates between this aspect of NeuRD and SPG. We show how CIX estimates can be used in a black-box reduction to construct bandit algorithms with regret bounds that hold with high probability and the benefits this entails for NeuRD-CIX in sequential decision-making settings. Our analysis reveals a bias--variance tradeoff between SPG and NeuRD, and shows how theory predicts that NeuRD-CIX will perform well more consistently than NeuRD while retaining NeuRD's advantages over SPG in non-stationary environments.

cs.LG↗

arXiv2022

Robust No-Regret Learning in Min-Max Stackelberg Games

The behavior of no-regret learning algorithms is well understood in two-player min-max (i.e, zero-sum) games. In this paper, we investigate the behavior of no-regret learning in min-max games with dependent strategy sets, where the strategy of the first player constrains the behavior of the second. Such games are best understood as sequential, i.e., min-max Stackelberg, games. We consider two settings, one in which only the first player chooses their actions using a no-regret algorithm while the second player best responds, and one in which both players use no-regret algorithms. For the former case, we show that no-regret dynamics converge to a Stackelberg equilibrium. For the latter case, we introduce a new type of regret, which we call Lagrangian regret, and show that if both players minimize their Lagrangian regrets, then play converges to a Stackelberg equilibrium. We then observe that online mirror descent (OMD) dynamics in these two settings correspond respectively to a known nested (i.e., sequential) gradient descent-ascent (GDA) algorithm and a new simultaneous GDA-like algorithm, thereby establishing convergence of these algorithms to Stackelberg equilibrium. Finally, we analyze the robustness of OMD dynamics to perturbations by investigating online min-max Stackelberg games. We prove that OMD dynamics are robust for a large class of online min-max games with independent strategy sets. In the dependent case, we demonstrate the robustness of OMD dynamics experimentally by simulating them in online Fisher markets, a canonical example of a min-max Stackelberg game with dependent strategy sets.

cs.GT↗

arXiv2022

A Consumer-Theoretic Characterization of Fisher Market Equilibria

In this paper, we bring consumer theory to bear in the analysis of Fisher markets whose buyers have arbitrary continuous, concave, homogeneous (CCH) utility functions representing locally non-satiated preferences. The main tools we use are the dual concepts of expenditure minimization and indirect utility maximization. First, we use expenditure functions to construct a new convex program whose dual, like the dual of the Eisenberg-Gale program, characterizes the equilibrium prices of CCH Fisher markets. We then prove that the subdifferential of the dual of our convex program is equal to the negative excess demand in the associated market, which makes generalized gradient descent equivalent to computing equilibrium prices via tâtonnement. Finally, we run a series of experiments which suggest that tâtonnement may converge at a rate of $O\left(\frac{(1+E)}{t^2}\right)$ in CCH Fisher markets that comprise buyers with elasticity of demand bounded by $E$. Our novel characterization of equilibrium prices may provide a path to proving the convergence of tâtonnement in Fisher markets beyond those in which buyers utilities exhibit constant elasticity of substitution.

cs.GT↗

arXiv2021

A Dynamic Programming Algorithm to Compute Joint Distribution of Order Statistics on Graphs

Order statistics play a fundamental role in statistical procedures such as risk estimation, outlier detection, and multiple hypothesis testing as well as in the analyses of mechanism design, queues, load balancing, and various other logistical processes involving ranks. In some of these cases, it may be desirable to compute the \textit{exact} values from the joint distribution of $d$ order statistics. While this problem is already computationally difficult even in the case of $n$ independent random variables, the random variables often have no such independence guarantees. Existing methods obtain the cumulative distribution indirectly by first computing and then aggregating over the marginal distributions. In this paper, we provide a more direct, efficient algorithm to compute cumulative joint order statistic distributions of dependent random variables that improves an existing dynamic programming solution via dimensionality reduction techniques. Our solution guarantees a $O(\frac{d^{d-1}}{n})$ and $O(d^{d})$ factor of improvement in both time and space complexity respectively over previous methods.

cs.DS↗

arXiv2021

Learning Competitive Equilibria in Noisy Combinatorial Markets

We present a methodology to robustly estimate the competitive equilibria (CE) of combinatorial markets under the assumption that buyers do not know their precise valuations for bundles of goods, but instead can only provide noisy estimates. We first show tight lower- and upper-bounds on the buyers' utility loss, and hence the set of CE, given a uniform approximation of one market by another. We then develop a learning framework for our setup, and present two probably-approximately-correct algorithms for learning CE, i.e., producing uniform approximations that preserve CE, with finite-sample guarantees. The first is a baseline that uses Hoeffding's inequality to produce a uniform approximation of buyers' valuations with high probability. The second leverages a connection between the first welfare theorem of economics and uniform approximations to adaptively prune value queries when it determines that they are provably not part of a CE. We experiment with our algorithms and find that the pruning algorithm achieves better estimates than the baseline with far fewer samples.

cs.GT↗

arXiv2019

Learning Equilibria of Simulation-Based Games

We tackle a fundamental problem in empirical game-theoretic analysis (EGTA), that of learning equilibria of simulation-based games. Such games cannot be described in analytical form; instead, a black-box simulator can be queried to obtain noisy samples of utilities. Our approach to EGTA is in the spirit of probably approximately correct learning. We design algorithms that learn so-called empirical games, which uniformly approximate the utilities of simulation-based games with finite-sample guarantees. These algorithms can be instantiated with various concentration inequalities. Building on earlier work, we first apply Hoeffding's bound, but as the size of the game grows, this bound eventually becomes statistically intractable; hence, we also use the Rademacher complexity. Our main results state: with high probability, all equilibria of the simulation-based game are approximate equilibria in the empirical game (perfect recall); and conversely, all approximate equilibria in the empirical game are approximate equilibria in the simulation-based game (approximately perfect precision). We evaluate our algorithms on several synthetic games, showing that they make frugal use of data, produce accurate estimates more often than the theory predicts, and are robust to different forms of noise.

cs.GT↗

arXiv2019

Artificial Intelligence for Social Good

The Computing Community Consortium (CCC), along with the White House Office of Science and Technology Policy (OSTP), and the Association for the Advancement of Artificial Intelligence (AAAI), co-sponsored a public workshop on Artificial Intelligence for Social Good on June 7th, 2016 in Washington, DC. This was one of five workshops that OSTP co-sponsored and held around the country to spur public dialogue on artificial intelligence, machine learning, and to identify challenges and opportunities related to AI. In the AI for Social Good workshop, the successful deployments and the potential use of AI in various topics that are essential for social good were discussed, including but not limited to urban computing, health, environmental sustainability, and public welfare. This report highlights each of these as well as a number of crosscutting issues.

cs.CY↗

arXiv2017

On Approximate Welfare- and Revenue-Maximizing Equilibria for Size-Interchangeable Bidders

In a Walrasian equilibrium (WE), all bidders are envy-free (EF), meaning that their allocation maximizes their utility; and the market clears (MC), meaning that the price of unallocated goods is zero. EF is desirable to ensure the long-term viability of the market. MC ensures that demand meets supply. Any allocation that is part of a WE is also welfare-maximizing; however, it need not be revenue-maximizing. Furthermore, WE need not exist, e.g., in markets where bidders have combinatorial valuations. The traditional approach to simultaneously addressing both existence and low revenue is to relax the MC condition and instead require the price of unallocated goods be some, positive reserve price. The resulting solution concept, known as Envy-Free Pricing (EFP), has been studied in some special cases, e.g., single-minded bidders. In this paper, we go one step further; we relax EF as well as MC. We propose a relaxation of the EF condition where only winners are envy-free, and further relax the MC condition so that unallocated goods are priced at least at the reserve. We call this new solution concept Restricted Envy-Free Pricing (REFP). We investigate what REFP entails for single-minded bidders, and show that for size-interchangeable bidders (a generalization of single-minded introduced in this paper) we can compute a REFP in polynomial time, given a fixed allocation. As in the study of EFP, we remain interested in maximizing seller revenue. Instead of computing an outcome that simultaneously yields an allocation and corresponding prices, one could first solve for an allocation that respects a reserve price, and then solve for a corresponding set of supporting prices, each one at least the reserve. This two-step process fails in the case of EFP since, given a fixed allocation, envy-free prices need not exist. However, restricted envy-free prices always exist...

cs.GT↗

arXiv2017

Simple vs Optimal Mechanisms in Auctions with Convex Payments

We investigate approximately optimal mechanisms in settings where bidders' utility functions are non-linear; specifically, convex, with respect to payments (such settings arise, for instance, in procurement auctions for energy). We provide constant factor approximation guarantees for mechanisms that are independent of bidders' private information (i.e., prior-free), and for mechanisms that rely to an increasing extent on that information (i.e., detail free). We also describe experiments, which show that for randomly drawn monotone hazard rate distributions, our mechanisms achieve at least 80\% of the optimal revenue, on average. Both our theoretical and experimental results show that in the convex payment setting, it is desirable to allocate across multiple bidders, rather than only to bidders with the highest (virtual) value, as in the traditional quasi-linear utility setting.

cs.GT↗

arXiv2016

Optimal Auctions with Convex Perceived Payments

Myerson derived a simple and elegant solution to the single-parameter revenue-maximization problem in his seminal work on optimal auction design assuming the usual model of quasi-linear utilities. In this paper, we consider a slight generalization of this usual model---from linear to convex "perceived" payments. This more general problem does not appear to admit a solution as simple and elegant as Myerson's. While some of Myerson's results extend to our setting, like his payment formula (suitably adjusted), others do not. For example, we observe that the solutions to the Bayesian and the robust (i.e., non-Bayesian) optimal auction design problems in the convex perceived payment setting do not coincide like they do in the case of linear payments. We therefore study the two problems in turn. We derive an upper and a heuristic lower bound on expected revenue in our setting. These bounds are easily computed pointwise, and yield monotonic allocation rules, so can be supported by Myerson payments (suitably adjusted). In this way, our bounds yield heuristics that approximate the optimal robust auction, assuming convex perceived payments. We close with experiments, the final set of which massages the output of one of the closed-form heuristics for the robust problem into an extremely fast, near-optimal heuristic solution to the Bayesian optimal auction design problem.

cs.GT↗

arXiv2014

RoxyBot-06: Stochastic Prediction and Optimization in TAC Travel

In this paper, we describe our autonomous bidding agent, RoxyBot, who emerged victorious in the travel division of the 2006 Trading Agent Competition in a photo finish. At a high level, the design of many successful trading agents can be summarized as follows: (i) price prediction: build a model of market prices; and (ii) optimization: solve for an approximately optimal set of bids, given this model. To predict, RoxyBot builds a stochastic model of market prices by simulating simultaneous ascending auctions. To optimize, RoxyBot relies on the sample average approximation method, a stochastic optimization technique.

cs.GT↗