Richard Combes

Richard Combes, Zwi Altman, Eitan Altman

Future communication networks are expected to feature autonomic (or self-organizing) mechanisms to ease deployment (self-configuration), tune parameters automatically (self-optimization) and repair the network (self-healing). Self-organizing mechanisms have been designed as stand-alone entities, even though multiple mechanisms will run in parallel in operational networks. An efficient coordination mechanism will be the major enabler for large scale deployment of self-organizing networks. We model self-organizing mechanisms as control loops, and study the conditions for stability when running control loops in parallel. Based on control theory and Lyapunov stability, we propose a coordination mechanism to stabilize the system, which can be implemented in a distributed fashion. The mechanism remains valid in the presence of measurement noise via stochastic approximation. Instability and coordination in the context of wireless networks are illustrated with two examples and the influence of network geometry is investigated. We are essentially concerned with linear systems, and the applicability of our results for non-linear systems is discussed.

Julien Floquet, Richard Combes, Zwi Altman

We consider hierarchical beamforming in wireless networks. For a given population of flows, we propose computationally efficient algorithms for fair rate allocation including proportional fairness and max-min fairness. We next propose closed-form formulas for flow level performance, for both elastic (with either proportional fairness and max-min fairness) and streaming traffic. We further assess the performance of hierarchical beamforming using numerical experiments. Since the proposed solutions have low complexity compared to conventional beamforming, our work suggests that hierarchical beamforming is a promising candidate for the implementation of beamforming in future cellular networks.

William Réveillard, Richard Combes

We consider a stochastic multi-armed bandit problem with i.i.d. rewards where the expected reward function is multimodal with at most m modes. We propose the first known computationally tractable algorithm for computing the solution to the Graves-Lai optimization problem, which in turn enables the implementation of asymptotically optimal algorithms for this bandit problem. The code for the proposed algorithms is publicly available at https://github.com/wilrev/MultimodalBandits

Raymond Zhang, Hédi Hadiji, Richard Combes

We consider the maximization of $x^\top θ$ over $(x,θ) \in \mathcal{X} \times Θ$, with $\mathcal{X} \subset \mathbb{R}^d$ convex and $Θ\subset \mathbb{R}^d$ an ellipsoid. This problem is fundamental in linear bandits, as the learner must solve it at every time step using optimistic algorithms. We first show that for some sets $\mathcal{X}$ e.g. $\ell_p$ balls with $p>2$, no efficient algorithms exist unless $\mathcal{P} = \mathcal{NP}$. We then provide two novel algorithms solving this problem efficiently when $\mathcal{X}$ is a centered ellipsoid. Our findings provide the first known method to implement optimistic algorithms for linear bandits in high dimensions.

Raymond Zhang, Hedi Hadiji, Richard Combes

We consider linear stochastic bandits where the set of actions is an ellipsoid. We provide the first known minimax optimal algorithm for this problem. We first derive a novel information-theoretic lower bound on the regret of any algorithm, which must be at least $Ω(\min(d σ\sqrt{T} + d \|θ\|_{A}, \|θ\|_{A} T))$ where $d$ is the dimension, $T$ the time horizon, $σ^2$ the noise variance, $A$ a matrix defining the set of actions and $θ$ the vector of unknown parameters. We then provide an algorithm whose regret matches this bound to a multiplicative universal constant. The algorithm is non-classical in the sense that it is not optimistic, and it is not a sampling algorithm. The main idea is to combine a novel sequential procedure to estimate $\|θ\|$, followed by an explore-and-commit strategy informed by this estimate. The algorithm is highly computationally efficient, and a run requires only time $O(dT + d^2 \log(T/d) + d^3)$ and memory $O(d^2)$, in contrast with known optimistic algorithms, which are not implementable in polynomial time. We go beyond minimax optimality and show that our algorithm is locally asymptotically minimax optimal, a much stronger notion of optimality. We further provide numerical experiments to illustrate our theoretical findings.

William Réveillard, Vasileios Saketos, Alexandre Proutiere et al.

We introduce and study an online problem arising in question answering systems. In this problem, an agent must sequentially classify user-submitted queries represented by $d$-dimensional embeddings drawn i.i.d. from an unknown distribution. The agent may consult a costly human expert for the correct label, or guess on her own without receiving feedback. The goal is to minimize regret against an oracle with free expert access. When the time horizon $T$ is at least exponential in the embedding dimension $d$, one can learn the geometry of the class regions: in this regime, we propose the Conservative Hull-based Classifier (CHC), which maintains convex hulls of expert-labeled queries and calls the expert as soon as a query lands outside all known hulls. CHC attains $\mathcal{O}(\log^d T)$ regret in $T$ and is minimax optimal for $d=1$. Otherwise, the geometry cannot be reliably learned without additional distributional assumptions. We show that when the queries are drawn from a subgaussian mixture, for $T \le e^d$, a Center-based Classifier (CC) achieves regret proportional to $N\log{N}$ where $N$ is the number of labels. To bridge these regimes, we introduce the Generalized Hull-based Classifier (GHC), a practical extension of CHC that allows for more aggressive guessing via a tunable threshold parameter. Our approach is validated with experiments, notably on real-world question-answering datasets using embeddings derived from state-of-the-art large language models.

Wei Huang, Richard Combes, Andrea Araldo et al.

We study the placement of virtual functions aimed at minimizing the cost. We propose a novel algorithm, using ideas based on multi-armed bandits. We prove that these algorithms learn the optimal placement policy rapidly, and their regret grows at a rate at most $O( N M \sqrt{T\ln T} )$ while respecting the feasibility constraints with high probability, where $T$ is total time slots, $M$ is the number of classes of function and $N$ is the number of computation nodes. We show through numerical experiments that the proposed algorithm both has good practical performance and modest computational complexity. We propose an acceleration technique that allows the algorithm to achieve good performance also in large networks where computational power is limited. Our experiments are fully reproducible, and the code is publicly available.

Wei Huang, Richard Combes, Cindy Trinh

We propose a novel algorithm for multi-player multi-armed bandits without collision sensing information. Our algorithm circumvents two problems shared by all state-of-the-art algorithms: it does not need as an input a lower bound on the minimal expected reward of an arm, and its performance does not scale inversely proportionally to the minimal expected reward. We prove a theoretical regret upper bound to justify these claims. We complement our theoretical results with numerical experiments, showing that the proposed algorithm outperforms state-of-the-art in practice as well.

Cindy Trinh, Richard Combes

Motivated by applications in cognitive radio networks, we consider the decentralized multi-player multi-armed bandit problem, without collision nor sensing information. We propose Randomized Selfish KL-UCB, an algorithm with very low computational complexity, inspired by the Selfish KL-UCB algorithm, which has been abandoned as it provably performs sub-optimally in some cases. We subject Randomized Selfish KL-UCB to extensive numerical experiments showing that it far outperforms state-of-the-art algorithms in almost all environments, sometimes by several orders of magnitude, and without the additional knowledge required by state-of-the-art algorithms. We also emphasize the potential of this algorithm for the more realistic dynamic setting, and support our claims with further experiments. We believe that the low complexity and high performance of Randomized Selfish KL-UCB makes it the most suitable for implementation in practical systems amongst known algorithms.

Thibaut Cuvelier, Richard Combes, Eric Gourdin

We consider combinatorial semi-bandits with uncorrelated Gaussian rewards. In this article, we propose the first method, to the best of our knowledge, that enables to compute the solution of the Graves-Lai optimization problem in polynomial time for many combinatorial structures of interest. In turn, this immediately yields the first known approach to implement asymptotically optimal algorithms in polynomial time for combinatorial semi-bandits.

Raymond Zhang, Richard Combes

In this paper we consider Thompson Sampling (TS) for combinatorial semi-bandits. We demonstrate that, perhaps surprisingly, TS is sub-optimal for this problem in the sense that its regret scales exponentially in the ambient dimension, and its minimax regret scales almost linearly. This phenomenon occurs under a wide variety of assumptions including both non-linear and linear reward functions, with Bernoulli distributed rewards and uniform priors. We also show that including a fixed amount of forced exploration to TS does not alleviate the problem. We complement our theoretical results with numerical results and show that in practice TS indeed can perform very poorly in some high dimensional situations.

Richard Combes, Mikael Touati

We consider the problem of solving random parity games. We prove that parity games exibit a phase transition threshold above $d_P$, so that when the degree of the graph that defines the game has a degree $d > d_P$ then there exists a polynomial time algorithm that solves the game with high probability when the number of nodes goes to infinity. We further propose the SWCP (Self-Winning Cycles Propagation) algorithm and show that, when the degree is large enough, SWCP solves the game with high probability. Furthermore, the complexity of SWCP is polynomial $O\Big(|{\cal V}|^2 + |{\cal V}||{\cal E}|\Big)$. The design of SWCP is based on the threshold for the appearance of particular types of cycles in the players' respective subgraphs. We further show that non-sparse games can be solved in time $O(|{\cal V}|)$ with high probability, and emit a conjecture concerning the hardness of the $d=2$ case.

Thibaut Cuvelier, Richard Combes, Eric Gourdin

We consider combinatorial semi-bandits over a set of arms ${\cal X} \subset \{0,1\}^d$ where rewards are uncorrelated across items. For this problem, the algorithm ESCB yields the smallest known regret bound $R(T) = {\cal O}\Big( {d (\ln m)^2 (\ln T) \over Δ_{\min} }\Big)$, but it has computational complexity ${\cal O}(|{\cal X}|)$ which is typically exponential in $d$, and cannot be used in large dimensions. We propose the first algorithm which is both computationally and statistically efficient for this problem with regret $R(T) = {\cal O} \Big({d (\ln m)^2 (\ln T)\over Δ_{\min} }\Big)$ and computational complexity ${\cal O}(T {\bf poly}(d))$. Our approach involves carefully designing an approximate version of ESCB with the same regret guarantees, showing that this approximate algorithm can be implemented in time ${\cal O}(T {\bf poly}(d))$ by repeatedly maximizing a linear function over ${\cal X}$ subject to a linear budget constraint, and showing how to solve this maximization problems efficiently.

Cindy Trinh, Emilie Kaufmann, Claire Vernade et al.

Stochastic Rank-One Bandits (Katarya et al, (2017a,b)) are a simple framework for regret minimization problems over rank-one matrices of arms. The initially proposed algorithms are proved to have logarithmic regret, but do not match the existing lower bound for this problem. We close this gap by first proving that rank-one bandits are a particular instance of unimodal bandits, and then providing a new analysis of Unimodal Thompson Sampling (UTS), initially proposed by Paladino et al (2017). We prove an asymptotically optimal regret bound on the frequentist regret of UTS and we support our claims with simulations showing the significant improvement of our method compared to the state-of-the-art.

Richard Combes, Mikael Touati

We consider the problem of estimating from sample paths the absolute spectral gap $γ_*$ of a reversible, irreducible and aperiodic Markov chain $(X_t)_{t \in \mathbb{N}}$ over a finite state space $Ω$. We propose the ${\tt UCPI}$ (Upper Confidence Power Iteration) algorithm for this problem, a low-complexity algorithm which estimates the spectral gap in time ${\cal O}(n)$ and memory space ${\cal O}((\ln n)^2)$ given $n$ samples. This is in stark contrast with most known methods which require at least memory space ${\cal O}(|Ω|)$, so that they cannot be applied to large state spaces. Furthermore, ${\tt UCPI}$ is amenable to parallel implementation.

Richard Combes, Stefan Magureanu, Alexandre Proutiere

This paper introduces and addresses a wide class of stochastic bandit problems where the function mapping the arm to the corresponding reward exhibits some known structural properties. Most existing structures (e.g. linear, Lipschitz, unimodal, combinatorial, dueling, ...) are covered by our framework. We derive an asymptotic instance-specific regret lower bound for these problems, and develop OSSB, an algorithm whose regret matches this fundamental limit. OSSB is not based on the classical principle of "optimism in the face of uncertainty" or on Thompson sampling, and rather aims at matching the minimal exploration rates of sub-optimal arms as characterized in the derivation of the regret lower bound. We illustrate the efficiency of OSSB using numerical experiments in the case of the linear bandit problem and show that OSSB outperforms existing algorithms, including Thompson sampling.

Jung-hun Kim, Se-Young Yun, Minchan Jeong et al.

We study contextual linear bandit problems under feature uncertainty, where the features are noisy and have missing entries. To address the challenges posed by this noise, we analyze Bayesian oracles given the observed noisy features. Our Bayesian analysis reveals that the optimal hypothesis can significantly deviate from the underlying realizability function, depending on the noise characteristics. These deviations are highly non-intuitive and do not occur in classical noiseless setups. This implies that classical approaches cannot guarantee a non-trivial regret bound. Therefore, we propose an algorithm that aims to approximate the Bayesian oracle based on the observed information under this model, achieving $\tilde{O}(d\sqrt{T})$ regret bound when there is a large number of arms. We demonstrate the proposed algorithm using synthetic and real-world datasets.

Thomas Bonald, Richard Combes

We consider the problem of accurately estimating the reliability of workers based on noisy labels they provide, which is a fundamental question in crowdsourcing. We propose a novel lower bound on the minimax estimation error which applies to any estimation procedure. We further propose Triangular Estimation (TE), an algorithm for estimating the reliability of workers. TE has low complexity, may be implemented in a streaming setting when labels are provided by workers in real time, and does not rely on an iterative procedure. We further prove that TE is minimax optimal and matches our lower bound. We conclude by assessing the performance of TE and other state-of-the-art algorithms on both synthetic and real-world data sets.

Thomas Bonald, Richard Combes

We propose a streaming algorithm for the binary classification of data based on crowdsourcing. The algorithm learns the competence of each labeller by comparing her labels to those of other labellers on the same tasks and uses this information to minimize the prediction error rate on each task. We provide performance guarantees of our algorithm for a fixed population of independent labellers. In particular, we show that our algorithm is optimal in the sense that the cumulative regret compared to the optimal decision with known labeller error probabilities is finite, independently of the number of tasks to label. The complexity of the algorithm is linear in the number of labellers and the number of tasks, up to some logarithmic factors. Numerical experiments illustrate the performance of our algorithm compared to existing algorithms, including simple majority voting and expectation-maximization algorithms, on both synthetic and real datasets.

Richard Combes

We generalize McDiarmid's inequality for functions with bounded differences on a high probability set, using an extension argument. Those functions concentrate around their conditional expectations. We further extend the results to concentration in general metric spaces.

Richard Combes, M. Sadegh Talebi, Alexandre Proutiere et al.

This paper investigates stochastic and adversarial combinatorial multi-armed bandit problems. In the stochastic setting under semi-bandit feedback, we derive a problem-specific regret lower bound, and discuss its scaling with the dimension of the decision space. We propose ESCB, an algorithm that efficiently exploits the structure of the problem and provide a finite-time analysis of its regret. ESCB has better performance guarantees than existing algorithms, and significantly outperforms these algorithms in practice. In the adversarial setting under bandit feedback, we propose \textsc{CombEXP}, an algorithm with the same regret scaling as state-of-the-art algorithms, but with lower computational complexity for some combinatorial problems.

Richard Combes, Alexandre Proutiere

We consider stochastic bandit problems with a continuous set of arms and where the expected reward is a continuous and unimodal function of the arm. No further assumption is made regarding the smoothness and the structure of the expected reward function. For these problems, we propose the Stochastic Pentachotomy (SP) algorithm, and derive finite-time upper bounds on its regret and optimization error. In particular, we show that, for any expected reward function $μ$ that behaves as $μ(x)=μ(x^\star)-C|x-x^\star|^ξ$ locally around its maximizer $x^\star$ for some $ξ, C>0$, the SP algorithm is order-optimal. Namely its regret and optimization error scale as $O(\sqrt{T\log(T)})$ and $O(\sqrt{\log(T)/T})$, respectively, when the time horizon $T$ grows large. These scalings are achieved without the knowledge of $ξ$ and $C$. Our algorithm is based on asymptotically optimal sequential statistical tests used to successively trim an interval that contains the best arm with high probability. To our knowledge, the SP algorithm constitutes the first sequential arm selection rule that achieves a regret and optimization error scaling as $O(\sqrt{T})$ and $O(1/\sqrt{T})$, respectively, up to a logarithmic factor for non-smooth expected reward functions, as well as for smooth functions with unknown smoothness.

Richard Combes, Alexandre Proutiere

We consider stochastic multi-armed bandits where the expected reward is a unimodal function over partially ordered arms. This important class of problems has been recently investigated in (Cope 2009, Yu 2011). The set of arms is either discrete, in which case arms correspond to the vertices of a finite graph whose structure represents similarity in rewards, or continuous, in which case arms belong to a bounded interval. For discrete unimodal bandits, we derive asymptotic lower bounds for the regret achieved under any algorithm, and propose OSUB, an algorithm whose regret matches this lower bound. Our algorithm optimally exploits the unimodal structure of the problem, and surprisingly, its asymptotic regret does not depend on the number of arms. We also provide a regret upper bound for OSUB in non-stationary environments where the expected rewards smoothly evolve over time. The analytical results are supported by numerical experiments showing that OSUB performs significantly better than the state-of-the-art algorithms. For continuous sets of arms, we provide a brief discussion. We show that combining an appropriate discretization of the set of arms with the UCB algorithm yields an order-optimal regret, and in practice, outperforms recently proposed algorithms designed to exploit the unimodal structure.

Stefan Magureanu, Richard Combes, Alexandre Proutiere

We consider stochastic multi-armed bandit problems where the expected reward is a Lipschitz function of the arm, and where the set of arms is either discrete or continuous. For discrete Lipschitz bandits, we derive asymptotic problem specific lower bounds for the regret satisfied by any algorithm, and propose OSLB and CKL-UCB, two algorithms that efficiently exploit the Lipschitz structure of the problem. In fact, we prove that OSLB is asymptotically optimal, as its asymptotic regret matches the lower bound. The regret analysis of our algorithms relies on a new concentration inequality for weighted sums of KL divergences between the empirical distributions of rewards and their true distributions. For continuous Lipschitz bandits, we propose to first discretize the action space, and then apply OSLB or CKL-UCB, algorithms that provably exploit the structure efficiently. This approach is shown, through numerical experiments, to significantly outperform existing algorithms that directly deal with the continuous set of arms. Finally the results and algorithms are extended to contextual bandits with similarities.

Richard Combes, Alexandre Proutiere

In this paper, we investigate dynamic channel and rate selection in cognitive radio systems which exploit a large number of channels free from primary users. In such systems, transmitters may rapidly change the selected (channel, rate) pair to opportunistically learn and track the pair offering the highest throughput. We formulate the problem of sequential channel and rate selection as an online optimization problem, and show its equivalence to a {\it structured} Multi-Armed Bandit problem. The structure stems from inherent properties of the achieved throughput as a function of the selected channel and rate. We derive fundamental performance limits satisfied by {\it any} channel and rate adaptation algorithm, and propose algorithms that achieve (or approach) these limits. In turn, the proposed algorithms optimally exploit the inherent structure of the throughput. We illustrate the efficiency of our algorithms using both test-bed and simulation experiments, in both stationary and non-stationary radio environments. In stationary environments, the packet successful transmission probabilities at the various channel and rate pairs do not evolve over time, whereas in non-stationary environments, they may evolve. In practical scenarios, the proposed algorithms are able to track the best channel and rate quite accurately without the need of any explicit measurement and feedback of the quality of the various channels.

M. Sadegh Talebi, Zhenhua Zou, Richard Combes et al.

This paper studies online shortest path routing over multi-hop networks. Link costs or delays are time-varying and modeled by independent and identically distributed random processes, whose parameters are initially unknown. The parameters, and hence the optimal path, can only be estimated by routing packets through the network and observing the realized delays. Our aim is to find a routing policy that minimizes the regret (the cumulative difference of expected delay) between the path chosen by the policy and the unknown optimal path. We formulate the problem as a combinatorial bandit optimization problem and consider several scenarios that differ in where routing decisions are made and in the information available when making the decisions. For each scenario, we derive a tight asymptotic lower bound on the regret that has to be satisfied by any online routing policy. These bounds help us to understand the performance improvements we can expect when (i) taking routing decisions at each hop rather than at the source only, and (ii) observing per-link delays rather than end-to-end path delays. In particular, we show that (i) is of no use while (ii) can have a spectacular impact. Three algorithms, with a trade-off between computational complexity and performance, are proposed. The regret upper bounds of these algorithms improve over those of the existing algorithms, and they significantly outperform state-of-the-art algorithms in numerical experiments.

Richard Combes, Ilham El Bouloumi, Stephane Senecal et al.

The purpose of this paper is to develop a self-optimized association algorithm based on PGRL (Policy Gradient Reinforcement Learning), which is both scalable, stable and robust. The term robust means that performance degradation in the learning phase should be forbidden or limited to predefined thresholds. The algorithm is model-free (as opposed to Value Iteration) and robust (as opposed to Q-Learning). The association problem is modeled as a Markov Decision Process (MDP). The policy space is parameterized. The parameterized family of policies is then used as expert knowledge for the PGRL. The PGRL converges towards a local optimum and the average cost decreases monotonically during the learning process. The properties of the solution make it a good candidate for practical implementation. Furthermore, the robustness property allows to use the PGRL algorithm in an "always-on" learning mode.