Abstract:We present novel reductions from sample compression schemes in multiclass classification, regression, and adversarially robust learning settings to binary sample compression schemes. Assuming we have a compression scheme for binary classes of size $f(d_\mathrm{VC})$, where $d_\mathrm{VC}$ is the VC dimension, then we have the following results: (1) If the binary compression scheme is a majority-vote or a stable compression scheme, then there exists a multiclass compression scheme of size $O(f(d_\mathrm{G}))$, where $d_\mathrm{G}$ is the graph dimension. Moreover, for general binary compression schemes, we obtain a compression of size $O(f(d_\mathrm{G})\log|Y|)$, where $Y$ is the label space. (2) If the binary compression scheme is a majority-vote or a stable compression scheme, then there exists an $\epsilon$-approximate compression scheme for regression over $[0,1]$-valued functions of size $O(f(d_\mathrm{P}))$, where $d_\mathrm{P}$ is the pseudo-dimension. For general binary compression schemes, we obtain a compression of size $O(f(d_\mathrm{P})\log(1/\epsilon))$. These results would have significant implications if the sample compression conjecture, which posits that any binary concept class with a finite VC dimension admits a binary compression scheme of size $O(d_\mathrm{VC})$, is resolved (Littlestone and Warmuth, 1986; Floyd and Warmuth, 1995; Warmuth, 2003). Our results would then extend the proof of the conjecture immediately to other settings. We establish similar results for adversarially robust learning and also provide an example of a concept class that is robustly learnable but has no bounded-size compression scheme, demonstrating that learnability is not equivalent to having a compression scheme independent of the sample size, unlike in binary classification, where compression of size $2^{O(d_\mathrm{VC})}$ is attainable (Moran and Yehudayoff, 2016).
Abstract:We study the fundamental problem of sequential probability assignment, also known as online learning with logarithmic loss, with respect to an arbitrary, possibly nonparametric hypothesis class. Our goal is to obtain a complexity measure for the hypothesis class that characterizes the minimax regret and to determine a general, minimax optimal algorithm. Notably, the sequential $\ell_{\infty}$ entropy, extensively studied in the literature (Rakhlin and Sridharan, 2015, Bilodeau et al., 2020, Wu et al., 2023), was shown to not characterize minimax risk in general. Inspired by the seminal work of Shtarkov (1987) and Rakhlin, Sridharan, and Tewari (2010), we introduce a novel complexity measure, the \emph{contextual Shtarkov sum}, corresponding to the Shtarkov sum after projection onto a multiary context tree, and show that the worst case log contextual Shtarkov sum equals the minimax regret. Using the contextual Shtarkov sum, we derive the minimax optimal strategy, dubbed \emph{contextual Normalized Maximum Likelihood} (cNML). Our results hold for sequential experts, beyond binary labels, which are settings rarely considered in prior work. To illustrate the utility of this characterization, we provide a short proof of a new regret upper bound in terms of sequential $\ell_{\infty}$ entropy, unifying and sharpening state-of-the-art bounds by Bilodeau et al. (2020) and Wu et al. (2023).
Abstract:In this work, we investigate the problem of adapting to the presence or absence of causal structure in multi-armed bandit problems. In addition to the usual reward signal, we assume the learner has access to additional variables, observed in each round after acting. When these variables $d$-separate the action from the reward, existing work in causal bandits demonstrates that one can achieve strictly better (minimax) rates of regret (Lu et al., 2020). Our goal is to adapt to this favorable "conditionally benign" structure, if it is present in the environment, while simultaneously recovering worst-case minimax regret, if it is not. Notably, the learner has no prior knowledge of whether the favorable structure holds. In this paper, we establish the Pareto optimal frontier of adaptive rates. We prove upper and matching lower bounds on the possible trade-offs in the performance of learning in conditionally benign and arbitrary environments, resolving an open question raised by Bilodeau et al. (2022). Furthermore, we are the first to obtain instance-dependent bounds for causal bandits, by reducing the problem to the linear bandit setting. Finally, we examine the common assumption that the marginal distributions of the post-action contexts are known and show that a nontrivial estimate is necessary for better-than-worst-case minimax rates.
Abstract:In this work, we investigate the interplay between memorization and learning in the context of \emph{stochastic convex optimization} (SCO). We define memorization via the information a learning algorithm reveals about its training data points. We then quantify this information using the framework of conditional mutual information (CMI) proposed by Steinke and Zakynthinou (2020). Our main result is a precise characterization of the tradeoff between the accuracy of a learning algorithm and its CMI, answering an open question posed by Livni (2023). We show that, in the $L^2$ Lipschitz--bounded setting and under strong convexity, every learner with an excess error $\varepsilon$ has CMI bounded below by $\Omega(1/\varepsilon^2)$ and $\Omega(1/\varepsilon)$, respectively. We further demonstrate the essential role of memorization in learning problems in SCO by designing an adversary capable of accurately identifying a significant fraction of the training samples in specific SCO problems. Finally, we enumerate several implications of our results, such as a limitation of generalization bounds based on CMI and the incompressibility of samples in SCO problems.
Abstract:While ERM suffices to attain near-optimal generalization error in the stochastic learning setting, this is not known to be the case in the online learning setting, where algorithms for general concept classes rely on computationally inefficient oracles such as the Standard Optimal Algorithm (SOA). In this work, we propose an algorithm for online binary classification setting that relies solely on ERM oracle calls, and show that it has finite regret in the realizable setting and sublinearly growing regret in the agnostic setting. We bound the regret in terms of the Littlestone and threshold dimensions of the underlying concept class. We obtain similar results for nonparametric games, where the ERM oracle can be interpreted as a best response oracle, finding the best response of a player to a given history of play of the other players. In this setting, we provide learning algorithms that only rely on best response oracles and converge to approximate-minimax equilibria in two-player zero-sum games and approximate coarse correlated equilibria in multi-player general-sum games, as long as the game has a bounded fat-threshold dimension. Our algorithms apply to both binary-valued and real-valued games and can be viewed as providing justification for the wide use of double oracle and multiple oracle algorithms in the practice of solving large games.
Abstract:In this work, we aim to characterize the statistical complexity of realizable regression both in the PAC learning setting and the online learning setting. Previous work had established the sufficiency of finiteness of the fat shattering dimension for PAC learnability and the necessity of finiteness of the scaled Natarajan dimension, but little progress had been made towards a more complete characterization since the work of Simon 1997 (SICOMP '97). To this end, we first introduce a minimax instance optimal learner for realizable regression and propose a novel dimension that both qualitatively and quantitatively characterizes which classes of real-valued predictors are learnable. We then identify a combinatorial dimension related to the Graph dimension that characterizes ERM learnability in the realizable setting. Finally, we establish a necessary condition for learnability based on a combinatorial dimension related to the DS dimension, and conjecture that it may also be sufficient in this context. Additionally, in the context of online learning we provide a dimension that characterizes the minimax instance optimal cumulative loss up to a constant factor and design an optimal online learner for realizable regression, thus resolving an open question raised by Daskalakis and Golowich in STOC '22.
Abstract:We study robustness to test-time adversarial attacks in the regression setting with $\ell_p$ losses and arbitrary perturbation sets. We address the question of which function classes are PAC learnable in this setting. We show that classes of finite fat-shattering dimension are learnable. Moreover, for convex function classes, they are even properly learnable. In contrast, some non-convex function classes provably require improper learning algorithms. We also discuss extensions to agnostic learning. Our main technique is based on a construction of an adversarially robust sample compression scheme of a size determined by the fat-shattering dimension.
Abstract:We consider a seller faced with buyers which have the ability to delay their decision, which we call patience. Each buyer's type is composed of value and patience, and it is sampled i.i.d. from a distribution. The seller, using posted prices, would like to maximize her revenue from selling to the buyer. Our main results are the following. $\bullet$ We formalize this setting and characterize the resulting Stackelberg equilibrium, where the seller first commits to her strategy and then the buyers best respond. $\bullet$ We show a separation between the best fixed price, the best pure strategy, which is a fixed sequence of prices, and the best mixed strategy, which is a distribution over price sequences. $\bullet$ We characterize both the optimal pure strategy of the seller and the buyer's best response strategy to any seller's mixed strategy. $\bullet$ We show how to compute efficiently the optimal pure strategy and give an algorithm for the optimal mixed strategy (which is exponential in the maximum patience). We then consider a learning setting, where the seller does not have access to the distribution over buyer's types. Our main results are the following. $\bullet$ We derive a sample complexity bound for the learning of an approximate optimal pure strategy, by computing the fat-shattering dimension of this setting. $\bullet$ We give a general sample complexity bound for the approximate optimal mixed strategy. $\bullet$ We consider an online setting and derive a vanishing regret bound with respect to both the optimal pure strategy and the optimal mixed strategy.
Abstract:We study the problem of semi-supervised learning of an adversarially-robust predictor in the PAC model, where the learner has access to both labeled and unlabeled examples. The sample complexity in semi-supervised learning has two parameters, the number of labeled examples and the number of unlabeled examples. We consider the complexity measures, $VC_U \leq dim_U \leq VC$ and $VC^*$, where $VC$ is the standard $VC$-dimension, $VC^*$ is its dual, and the other two measures appeared in Montasser et al. (2019). The best sample bound known for robust supervised PAC learning is $O(VC \cdot VC^*)$, and we will compare our sample bounds to $\Lambda$ which is the minimal number of labeled examples required by any robust supervised PAC learning algorithm. Our main results are the following: (1) in the realizable setting it is sufficient to have $O(VC_U)$ labeled examples and $O(\Lambda)$ unlabeled examples. (2) In the agnostic setting, let $\eta$ be the minimal agnostic error. The sample complexity depends on the resulting error rate. If we allow an error of $2\eta+\epsilon$, it is still sufficient to have $O(VC_U)$ labeled examples and $O(\Lambda)$ unlabeled examples. If we insist on having an error $\eta+\epsilon$ then $\Omega(dim_U)$ labeled examples are necessary, as in the supervised case. The above results show that there is a significant benefit in semi-supervised robust learning, as there are hypothesis classes with $VC_U=0$ and $dim_U$ arbitrary large. In supervised learning, having access only to labeled examples requires at least $\Lambda \geq dim_U$ labeled examples. Semi-supervised require only $O(1)$ labeled examples and $O(\Lambda)$ unlabeled examples. A byproduct of our result is that if we assume that the distribution is robustly realizable by a hypothesis class, then with respect to the 0-1 loss we can learn with only $O(VC_U)$ labeled examples, even if the $VC$ is infinite.
Abstract:We provide improved estimates on the fat-shattering dimension of the $k$-fold maximum of real-valued function classes. The latter consists of all ways of choosing $k$ functions, one from each of the $k$ classes, and computing their pointwise maximum. The bound is stated in terms of the fat-shattering dimensions of the component classes. For linear and affine function classes, we provide a considerably sharper upper bound and a matching lower bound, achieving, in particular, an optimal dependence on $k$. Along the way, we point out and correct a number of erroneous claims in the literature.