Dirk van der Hoeven

Dirk van der Hoeven, Nikita Zhivotovskiy, Nicolò Cesa-Bianchi · eth-zurich

We consider prediction with expert advice for strongly convex and bounded losses, and investigate trade-offs between regret and "variance" (i.e., squared difference of learner's predictions and best expert predictions). With $K$ experts, the Exponentially Weighted Average (EWA) algorithm is known to achieve $O(\log K)$ regret. We prove that a variant of EWA either achieves a negative regret (i.e., the algorithm outperforms the best expert), or guarantees a $O(\log K)$ bound on both variance and regret. Building on this result, we show several examples of how variance of predictions can be exploited in learning. In the online to batch analysis, we show that a large empirical variance allows to stop the online to batch conversion early and outperform the risk of the best predictor in the class. We also recover the optimal rate of model selection aggregation when we do not consider early stopping. In online prediction with corrupted losses, we show that the effect of corruption on the regret can be compensated by a large variance. In online selective sampling, we design an algorithm that samples less when the variance is large, while guaranteeing the optimal regret bound in expectation. In online learning with abstention, we use a similar term as the variance to derive the first high-probability $O(\log K)$ regret bound in this setting. Finally, we extend our results to the setting of online linear regression.

Dirk van der Hoeven, Nikita Zhivotovskiy, Nicolò Cesa-Bianchi · eth-zurich

Online learning methods yield sequential regret bounds under minimal assumptions and provide in-expectation risk bounds for statistical learning. However, despite the apparent advantage of online guarantees over their statistical counterparts, recent findings indicate that in many important cases, regret bounds may not guarantee tight high-probability risk bounds in the statistical setting. In this work we show that online to batch conversions applied to general online learning algorithms can bypass this limitation. Via a general second-order correction to the loss function defining the regret, we obtain nearly optimal high-probability risk bounds for several classical statistical estimation problems, such as discrete distribution estimation, linear regression, logistic regression, and conditional density estimation. Our analysis relies on the fact that many online learning algorithms are improper, as they are not restricted to use predictors from a given reference class. The improper nature of our estimators enables significant improvements in the dependencies on various problem parameters. Finally, we discuss some computational advantages of our sequential algorithms over their existing batch counterparts.

Chloé Rouyer, Dirk van der Hoeven, Nicolò Cesa-Bianchi et al.

We consider online learning with feedback graphs, a sequential decision-making framework where the learner's feedback is determined by a directed graph over the action set. We present a computationally efficient algorithm for learning in this framework that simultaneously achieves near-optimal regret bounds in both stochastic and adversarial environments. The bound against oblivious adversaries is $\tilde{O} (\sqrt{αT})$, where $T$ is the time horizon and $α$ is the independence number of the feedback graph. The bound against stochastic environments is $O\big( (\ln T)^2 \max_{S\in \mathcal I(G)} \sum_{i \in S} Δ_i^{-1}\big)$ where $\mathcal I(G)$ is the family of all independent sets in a suitably defined undirected version of the graph and $Δ_i$ are the suboptimality gaps. The algorithm combines ideas from the EXP3++ algorithm for stochastic and adversarial bandits and the EXP3.G algorithm for feedback graphs with a novel exploration scheme. The scheme, which exploits the structure of the graph to reduce exploration, is key to obtain best-of-both-worlds guarantees with feedback graphs. We also extend our algorithm and results to a setting where the feedback graphs are allowed to change over time.

Emmanuel Esposito, Federico Fusco, Dirk van der Hoeven et al.

The framework of feedback graphs is a generalization of sequential decision-making with bandit or full information feedback. In this work, we study an extension where the directed feedback graph is stochastic, following a distribution similar to the classical Erdős-Rényi model. Specifically, in each round every edge in the graph is either realized or not with a distinct probability for each edge. We prove nearly optimal regret bounds of order $\min\bigl\{\min_{\varepsilon} \sqrt{(α_\varepsilon/\varepsilon) T},\, \min_{\varepsilon} (δ_\varepsilon/\varepsilon)^{1/3} T^{2/3}\bigr\}$ (ignoring logarithmic factors), where $α_{\varepsilon}$ and $δ_{\varepsilon}$ are graph-theoretic quantities measured on the support of the stochastic feedback graph $\mathcal{G}$ with edge probabilities thresholded at $\varepsilon$. Our result, which holds without any preliminary knowledge about $\mathcal{G}$, requires the learner to observe only the realized out-neighborhood of the chosen action. When the learner is allowed to observe the realization of the entire graph (but only the losses in the out-neighborhood of the chosen action), we derive a more efficient algorithm featuring a dependence on weighted versions of the independence and weak domination numbers that exhibits improved bounds for some special cases.

Dirk van der Hoeven, Ciara Pike-Burke, Hao Qiu et al.

We investigate online classification with paid stochastic experts. Here, before making their prediction, each expert must be paid. The amount that we pay each expert directly influences the accuracy of their prediction through some unknown Lipschitz "productivity" function. In each round, the learner must decide how much to pay each expert and then make a prediction. They incur a cost equal to a weighted sum of the prediction error and upfront payments for all experts. We introduce an online learning algorithm whose total cost after $T$ rounds exceeds that of a predictor which knows the productivity of all experts in advance by at most $\mathcal{O}(K^2(\log T)\sqrt{T})$ where $K$ is the number of experts. In order to achieve this result, we combine Lipschitz bandits and online classification with surrogate losses. These tools allow us to improve upon the bound of order $T^{2/3}$ one would obtain in the standard Lipschitz bandit setting. Our algorithm is empirically evaluated on synthetic data

Dirk van der Hoeven, Julia Olkhovskaia, Tim van Erven

We consider the fundamental problem of estimating a discrete distribution on a domain of size~$K$ with high probability in Kullback-Leibler divergence. We provide upper and lower bounds on the minimax estimation rate, which show that the optimal rate is between $\big(K + \ln(K)\ln(1/δ)\big) /n$ and $\big(K\ln\ln(K) + \ln(K)\ln(1/δ)\big) /n$ at error probability $δ$ and sample size $n$, which pins down the rate up to the doubly logarithmic factor $\ln \ln K$ that multiplies $K$. Our upper bound uses techniques from online learning to construct a novel estimator via online-to-batch conversion. Perhaps surprisingly, the tail behavior of the minimax rate is worse than for the squared total variation and squared Hellinger distance, for which it is $\big(K + \ln(1/δ)\big) /n$, i.e.\ without the $\ln K$ multiplying $\ln (1/δ)$. As a consequence, we cannot obtain a fully tight lower bound from the usual reduction to these smaller distances. Moreover, we show that this lower bound cannot be achieved by the standard lower bound approach based on a reduction to hypothesis testing, and instead we need to introduce a new reduction to what we call weak hypothesis testing. We investigate the source of the gap with other divergences further in refined results, which show that the total variation rate is achievable for Kullback-Leibler divergence after all (in fact by he maximum likelihood estimator) if we rule out outcome probabilities smaller than $O(\ln(K/δ) / n)$, which is a vanishing set as $n$ increases for fixed $K$ and~$δ$. This explains why minimax Kullback-Leibler estimation is more difficult than asymptotic estimation.

Antoine Moulin, Emmanuel Esposito, Dirk van der Hoeven

We consider the problem setting of prediction with expert advice with possibly heavy-tailed losses, i.e.\ the only assumption on the losses is an upper bound on their second moments, denoted by $θ$. We develop adaptive algorithms that do not require any prior knowledge about the range or the second moment of the losses. Existing adaptive algorithms have what is typically considered a lower-order term in their regret guarantees. We show that this lower-order term, which is often the maximum of the losses, can actually dominate the regret bound in our setting. Specifically, we show that even with small constant $θ$, this lower-order term can scale as $\sqrt{KT}$, where $K$ is the number of experts and $T$ is the time horizon. We propose adaptive algorithms with improved regret bounds that avoid the dependence on such a lower-order term and guarantee $\mathcal{O}(\sqrt{θT\log(K)})$ regret in the worst case, and $\mathcal{O}(θ\log(KT)/Δ_{\min})$ regret when the losses are sampled i.i.d.\ from some fixed distribution, where $Δ_{\min}$ is the difference between the mean losses of the second best expert and the best expert. Additionally, when the loss function is the squared loss, our algorithm also guarantees improved regret bounds over prior results.

Hidde Fokkema, Dirk van der Hoeven, Tor Lattimore et al.

We introduce a computationally efficient algorithm for zeroth-order bandit convex optimisation and prove that in the adversarial setting its regret is at most $d^{3.5} \sqrt{n} \mathrm{polylog}(n, d)$ with high probability where $d$ is the dimension and $n$ is the time horizon. In the stochastic setting the bound improves to $M d^{2} \sqrt{n} \mathrm{polylog}(n, d)$ where $M \in [d^{-1/2}, d^{-1 / 4}]$ is a constant that depends on the geometry of the constraint set and the desired computational properties.

Emmanuel Esposito, Saeed Masoudian, Hao Qiu et al.

We study a $K$-armed bandit with delayed feedback and intermediate observations. We consider a model where intermediate observations have a form of a finite state, which is observed immediately after taking an action, whereas the loss is observed after an adversarially chosen delay. We show that the regime of the mapping of states to losses determines the complexity of the problem, irrespective of whether the mapping of actions to states is stochastic or adversarial. If the mapping of states to losses is adversarial, then the regret rate is of order $\sqrt{(K+d)T}$ (within log factors), where $T$ is the time horizon and $d$ is a fixed delay. This matches the regret rate of a $K$-armed bandit with delayed feedback and without intermediate observations, implying that intermediate observations are not helpful. However, if the mapping of states to losses is stochastic, we show that the regret grows at a rate of $\sqrt{\big(K+\min\{|\mathcal{S}|,d\}\big)T}$ (within log factors), implying that if the number $|\mathcal{S}|$ of states is smaller than the delay, then intermediate observations help. We also provide refined high-probability regret upper bounds for non-uniform delays, together with experimental validation of our algorithms.

Dirk van der Hoeven, Lukas Zierahn, Tal Lancewicki et al.

We derive a new analysis of Follow The Regularized Leader (FTRL) for online learning with delayed bandit feedback. By separating the cost of delayed feedback from that of bandit feedback, our analysis allows us to obtain new results in three important settings. On the one hand, we derive the first optimal (up to logarithmic factors) regret bounds for combinatorial semi-bandits with delay and adversarial Markov decision processes with delay (and known transition functions). On the other hand, we use our analysis to derive an efficient algorithm for linear bandits with delay achieving near-optimal regret bounds. Our novel regret decomposition shows that FTRL remains stable across multiple rounds under mild assumptions on the Hessian of the regularizer.

Dirk van der Hoeven, Nicolò Cesa-Bianchi

We study nonstochastic bandits and experts in a delayed setting where delays depend on both time and arms. While the setting in which delays only depend on time has been extensively studied, the arm-dependent delay setting better captures real-world applications at the cost of introducing new technical challenges. In the full information (experts) setting, we design an algorithm with a first-order regret bound that reveals an interesting trade-off between delays and losses. We prove a similar first-order regret bound also for the bandit setting, when the learner is allowed to observe how many losses are missing. These are the first bounds in the delayed setting that depend on the losses and delays of the best arm only. When in the bandit setting no information other than the losses is observed, we still manage to prove a regret bound through a modification to the algorithm of Zimmert and Seldin (2020). Our analyses hinge on a novel bound on the drift, measuring how much better an algorithm can perform when given a look-ahead of one round.

Dirk van der Hoeven, Federico Fusco, Nicolò Cesa-Bianchi

We study the problem of online multiclass classification in a setting where the learner's feedback is determined by an arbitrary directed graph. While including bandit feedback as a special case, feedback graphs allow a much richer set of applications, including filtering and label efficient classification. We introduce Gappletron, the first online multiclass algorithm that works with arbitrary feedback graphs. For this new algorithm, we prove surrogate regret bounds that hold, both in expectation and with high probability, for a large class of surrogate losses. Our bounds are of order $B\sqrt{ρKT}$, where $B$ is the diameter of the prediction space, $K$ is the number of classes, $T$ is the time horizon, and $ρ$ is the domination number (a graph-theoretic parameter affecting the amount of exploration). In the full information case, we show that Gappletron achieves a constant surrogate regret of order $B^2K$. We also prove a general lower bound of order $\max\big\{B^2K,\sqrt{T}\big\}$ showing that our upper bounds are not significantly improvable. Experiments on synthetic data show that for various feedback graphs, our algorithm is competitive against known baselines.

Dirk van der Hoeven, Hédi Hadiji, Tim van Erven

We consider distributed online learning for joint regret with communication constraints. In this setting, there are multiple agents that are connected in a graph. Each round, an adversary first activates one of the agents to issue a prediction and provides a corresponding gradient, and then the agents are allowed to send a $b$-bit message to their neighbors in the graph. All agents cooperate to control the joint regret, which is the sum of the losses of the activated agents minus the losses evaluated at the best fixed common comparator parameters $u$. We observe that it is suboptimal for agents to wait for gradients that take too long to arrive. Instead, the graph should be partitioned into local clusters that communicate among themselves. Our main result is a new method that can adapt to the optimal graph partition for the adversarial activations and gradients, where the graph partition is selected from a set of candidate partitions. A crucial building block along the way is a new algorithm for online convex optimization with delayed gradient information that is comparator-adaptive, meaning that its joint regret scales with the norm of the comparator $||u||$. We further provide near-optimal gradient compression schemes depending on the ratio of $b$ and the dimension times the diameter of the graph.

Tim van Erven, Wouter M. Koolen, Dirk van der Hoeven

We provide a new adaptive method for online convex optimization, MetaGrad, that is robust to general convex losses but achieves faster rates for a broad class of special functions, including exp-concave and strongly convex functions, but also various types of stochastic and non-stochastic functions without any curvature. We prove this by drawing a connection to the Bernstein condition, which is known to imply fast rates in offline statistical learning. MetaGrad further adapts automatically to the size of the gradients. Its main feature is that it simultaneously considers multiple learning rates, which are weighted directly proportional to their empirical performance on the data using a new meta-algorithm. We provide three versions of MetaGrad. The full matrix version maintains a full covariance matrix and is applicable to learning tasks for which we can afford update time quadratic in the dimension. The other two versions provide speed-ups for high-dimensional learning tasks with an update time that is linear in the dimension: one is based on sketching, the other on running a separate copy of the basic algorithm per coordinate. We evaluate all versions of MetaGrad on benchmark online classification and regression tasks, on which they consistently outperform both online gradient descent and AdaGrad.

Dirk van der Hoeven

We present Gaptron, a randomized first-order algorithm for online multiclass classification. In the full information setting we show expected mistake bounds with respect to the logistic loss, hinge loss, and the smooth hinge loss with constant regret, where the expectation is with respect to the learner's randomness. In the bandit classification setting we show that Gaptron is the first linear time algorithm with $O(K\sqrt{T})$ expected regret, where $K$ is the number of classes. Additionally, the expected mistake bound of Gaptron does not depend on the dimension of the feature vector, contrary to previous algorithms with $O(K\sqrt{T})$ regret in the bandit classification setting. We present a new proof technique that exploits the gap between the zero-one loss and surrogate losses rather than exploiting properties such as exp-concavity or mixability, which are traditionally used to prove logarithmic or constant regret bounds.

Dirk van der Hoeven, Ashok Cutkosky, Haipeng Luo

We study bandit convex optimization methods that adapt to the norm of the comparator, a topic that has only been studied before for its full-information counterpart. Specifically, we develop convex bandit algorithms with regret bounds that are small whenever the norm of the comparator is small. We first use techniques from the full-information setting to develop comparator-adaptive algorithms for linear bandits. Then, we extend the ideas to convex bandits with Lipschitz or smooth loss functions, using a new single-point gradient estimator and carefully designed surrogate losses.

Dirk van der Hoeven, Tim van Erven, Wojciech Kotłowski

A standard introduction to online learning might place Online Gradient Descent at its center and then proceed to develop generalizations and extensions like Online Mirror Descent and second-order methods. Here we explore the alternative approach of putting Exponential Weights (EW) first. We show that many standard methods and their regret bounds then follow as a special case by plugging in suitable surrogate losses and playing the EW posterior mean. For instance, we easily recover Online Gradient Descent by using EW with a Gaussian prior on linearized losses, and, more generally, all instances of Online Mirror Descent based on regular Bregman divergences also correspond to EW with a prior that depends on the mirror map. Furthermore, appropriate quadratic surrogate losses naturally give rise to Online Gradient Descent for strongly convex losses and to Online Newton Step. We further interpret several recent adaptive methods (iProd, Squint, and a variation of Coin Betting for experts) as a series of closely related reductions to exp-concave surrogate losses that are then handled by Exponential Weights. Finally, a benefit of our EW interpretation is that it opens up the possibility of sampling from the EW posterior distribution instead of playing the mean. As already observed by Bubeck and Eldan, this recovers the best-known rate in Online Bandit Linear Optimization.