Note: When clicking on a Digital Object Identifier (DOI) number, you will be taken to an external site maintained by the publisher.
Some full text articles may not yet be available without a charge during the embargo (administrative interval).
What is a DOI Number?
Some links on this page may take you to nonfederal websites. Their policies may differ from this site.

The use of minmax optimization in the adversarial training of deep neural network classifiers, and the training of generative adversarial networks has motivated the study of nonconvexnonconcave optimization objectives, which frequently arise in these applications. Unfortunately, recent results have established that even approximate firstorder stationary points of such objectives are intractable, even under smoothness conditions, motivating the study of minmax objectives with additional structure. We introduce a new class of structured nonconvexnonconcave minmax optimization problems, proposing a generalization of the extragradient algorithm which provably converges to a stationary point. The algorithm applies not only to Euclidean spaces, but also to general ℓ𝑝normed finitedimensional real vector spaces. We also discuss its stability under stochastic oracles and provide bounds on its sample complexity. Our iteration complexity and sample complexity bounds either match or improve the best known bounds for the same or less general nonconvexnonconcave settings, such as those that satisfy variational coherence or in which a weak solution to the associated variational inequality problem is assumed to exist.

We consider the classical problem of selling a single item to a single bidder whose value for the item is drawn from a regular distribution F, in a "datapoor'' regime where Fis not known to the seller, and very few samples from Fare available. Prior work [Dhangwatnotai et al '10] has shown that one sample from Fcan be used to attain a 1/2factor approximation to the optimal revenue, but it has been challenging to improve this guarantee when more samples from Fare provided, even when two samples from Fare provided. In this case, the best approximation known to date is 0.509, achieved by the Empirical Revenue Maximizing (ERM) mechanism Babaioff et al. '18]. We improve this guarantee to 0.558, and provide a lower bound of 0.65. Our results are based on a general framework, based on factorrevealing Semidefinite Programming relaxations aiming to capture as tight as possible a superset of product measures of regular distributions, the challenge being that neither regularity constraints nor product measures are convex constraints. The framework is general and can be applied in more abstract settings to evaluate the performance of a policy chosen using independent samples from a distribution and applied on a fresh samplemore »

https://arxiv.org/abs/2010.13724 We study the question of obtaining lastiterate convergence rates for noregret learning algorithms in multiplayer games. We show that the optimistic gradient (OG) algorithm with a constant stepsize, which is noregret, achieves a lastiterate rate of O(1/T‾‾√) with respect to the gap function in smooth monotone games. This result addresses a question of Mertikopoulos & Zhou (2018), who asked whether extragradient approaches (such as OG) can be applied to achieve improved guarantees in the multiagent learning setting. The proof of our upper bound uses a new technique centered around an adaptive choice of potential function at each iteration. We also show that the O(1/T‾‾√) rate is tight for all pSCLI algorithms, which includes OG as a special case. As a byproduct of our lower bound analysis we additionally present a proof of a conjecture of Arjevani et al. (2015) which is more direct than previous approaches.

We obtain global, nonasymptotic convergence guarantees for independent learning algorithms in competitive reinforcement learning settings with two agents (i.e., zerosum stochastic games). We consider an episodic setting where in each episode, each player independently selects a policy and observes only their own actions and rewards, along with the state. We show that if both players run policy gradient methods in tandem, their policies will converge to a minmax equilibrium of the game, as long as their learning rates follow a twotimescale rule (which is necessary). To the best of our knowledge, this constitutes the first finitesample convergence result for independent policy gradient methods in competitive RL; prior work has largely focused on centralized, coordinated procedures for equilibrium computation.

Generative neural networks have been empirically found very promising in providing effective structural priors for compressed sensing, since they can be trained to span lowdimensional data manifolds in highdimensional signal spaces. Despite the nonconvexity of the resulting optimization problem, it has also been shown theoretically that, for neural networks with random Gaussian weights, a signal in the range of the network can be efficiently, approximately recovered from a few noisy measurements. However, a major bottleneck of these theoretical guarantees is a network expansivity condition: that each layer of the neural network must be larger than the previous by a logarithmic factor. Our main contribution is to break this strong expansivity assumption, showing that constant expansivity suffices to get efficient recovery algorithms, besides it also being informationtheoretically necessary. To overcome the theoretical bottleneck in existing approaches we prove a novel uniform concentration theorem for random functions that might not be Lipschitz but satisfy a relaxed notion which we call "pseudoLipschitzness." Using this theorem we can show that a matrix concentration inequality known as the Weight Distribution Condition (WDC), which was previously only known to hold for Gaussian matrices with logarithmic aspect ratio, in fact holds for constant aspect ratios too. Sincemore »

https://arxiv.org/abs/2007.14539 As in standard linear regression, in truncated linear regression, we are given access to observations (Ai,yi)i whose dependent variable equals yi=ATi⋅x∗+ηi, where x∗ is some fixed unknown vector of interest and ηi is independent noise; except we are only given an observation if its dependent variable yi lies in some "truncation set" S⊂ℝ. The goal is to recover x∗ under some favorable conditions on the Ai's and the noise distribution. We prove that there exists a computationally and statistically efficient method for recovering ksparse ndimensional vectors x∗ from m truncated samples, which attains an optimal ℓ2 reconstruction error of O((klogn)/m‾‾‾‾‾‾‾‾‾‾√). As a corollary, our guarantees imply a computationally efficient and informationtheoretically optimal algorithm for compressed sensing with truncation, which may arise from measurement saturation effects. Our result follows from a statistical and computational analysis of the Stochastic Gradient Descent (SGD) algorithm for solving a natural adaptation of the LASSO optimization problem that accommodates truncation. This generalizes the works of both: (1) [Daskalakis et al. 2018], where no regularization is needed due to the lowdimensionality of the data, and (2) [Wainright 2009], where the objective function is simple due to the absence of truncation. In order to deal with bothmore »

In this paper we study the smooth convexconcave saddle point problem. Specifically, we analyze the last iterate convergence properties of the Extragradient (EG) algorithm. It is well known that the ergodic (averaged) iterates of EG converge at a rate of O(1/T) (Nemirovski, 2004). In this paper, we show that the last iterate of EG converges at a rate of O(1/T‾‾√). To the best of our knowledge, this is the first paper to provide a convergence rate guarantee for the last iterate of EG for the smooth convexconcave saddle point problem. Moreover, we show that this rate is tight by proving a lower bound of Ω(1/T‾‾√) for the last iterate. This lower bound therefore shows a quadratic separation of the convergence rates of ergodic and last iterates in smooth convexconcave saddle point problems.

We identify the first static credible mechanism for multiitem additive auctions that achieves a constant factor of the optimal revenue. This is one instance of a more general framework for designing twopart tariff auctions, adapting the duality framework of Cai et al [CDW16]. Given a (not necessarily incentive compatible) auction format A satisfying certain technical conditions, our framework augments the auction with a personalized entry fee for each bidder, which must be paid before the auction can be accessed. These entry fees depend only on the prior distribution of bidder types, and in particular are independent of realized bids. Our framework can be used with many common auction formats, such as simultaneous firstprice, simultaneous secondprice, and simultaneous allpay auctions. If allpay auctions are used, we prove that the resulting mechanism is credible in the sense that the auctioneer cannot benefit by deviating from the stated mechanism after observing agent bids. If secondprice auctions are used, we obtain a truthful O(1)approximate mechanism with fixed entry fees that are amenable to tuning via online learning techniques. Our results for first price and allpay are the first revenue guarantees of nontruthful mechanisms in multidimensional environments; an open question in the literature [RST17].

Motivated by applications in Game Theory, Optimization, and Generative Adversarial Networks, recent work of Daskalakis et al \cite{DISZ17} and followup work of Liang and Stokes \cite{LiangS18} have established that a variant of the widely used Gradient Descent/Ascent procedure, called "Optimistic Gradient Descent/Ascent (OGDA)", exhibits lastiterate convergence to saddle points in {\em unconstrained} convexconcave minmax optimization problems. We show that the same holds true in the more general problem of {\em constrained} minmax optimization under a variant of the noregret MultiplicativeWeightsUpdate method called "Optimistic MultiplicativeWeights Update (OMWU)". This answers an open question of Syrgkanis et al \cite{SALS15}. The proof of our result requires fundamentally different techniques from those that exist in noregret learning literature and the aforementioned papers. We show that OMWU monotonically improves the KullbackLeibler divergence of the current iterate to the (appropriately normalized) minmax solution until it enters a neighborhood of the solution. Inside that neighborhood we show that OMWU is locally (asymptotically) stable converging to the exact solution. We believe that our techniques will be useful in the analysis of the last iterate of other learning algorithms.