publications

2024

1. NeurIPS

   This Too Shall Pass: Removing Stale Observations in Dynamic Bayesian Optimization

   Anthony Bardou, Patrick Thiran, and Giovanni Ranieri

   In NeurIPS’24: Thirty-Eighth Annual Conference on Neural Information Processing Systems 2024

   Abs Bib PDF Video

   Bayesian Optimization (BO) has proven to be very successful at optimizing a static, noisy, costly-to-evaluate black-box function f : \mathcalS \to \mathbbR. However, optimizing a black-box which is also a function of time (i.e., a dynamic function) f : \mathcalS \times \mathcalT \to \mathbbR remains a challenge, since a dynamic Bayesian Optimization (DBO) algorithm has to keep track of the optimum over time. This changes the nature of the optimization problem in at least three aspects: (i) querying an arbitrary point in \mathcalS \times \mathcalT is impossible, (ii) past observations become less and less relevant for keeping track of the optimum as time goes by and (iii) the DBO algorithm must have a high sampling frequency so it can collect enough relevant observations to keep track of the optimum through time. In this paper, we design a Wasserstein distance-based criterion able to quantify the relevancy of an observation with respect to future predictions. Then, we leverage this criterion to build W-DBO, a DBO algorithm able to remove irrelevant observations from its dataset on the fly, thus maintaining simultaneously a good predictive performance and a high sampling frequency, even in continuous-time optimization tasks with unknown horizon. Numerical experiments establish the superiority of W-DBO, which outperforms state-of-the-art methods by a comfortable margin.

   @inproceedings{wdbo,
     author = {Bardou, Anthony and Thiran, Patrick and Ranieri, Giovanni},
     booktitle = {NeurIPS'24: Thirty-Eighth Annual Conference on Neural Information Processing Systems},
     title = {This Too Shall Pass: Removing Stale Observations in Dynamic Bayesian Optimization},
     year = {2024},
     video = {PermSix-Hump_Camel_25.0_240.mp4},
   }

2. ICLR

   Relaxing the Additivity Constraints in Decentralized No-Regret High-Dimensional Bayesian Optimization

   Anthony Bardou, Patrick Thiran, and Thomas Begin

   In ICLR ’24: Int’l Conference on Learning Representations 2024

   Abs Bib PDF

   Bayesian Optimization (BO) is typically used to optimize an unknown function f that is noisy and costly to evaluate, by exploiting an acquisition function that must be maximized at each optimization step. Even if provably asymptotically optimal BO algorithms are efficient at optimizing low-dimensional functions, scaling them to high-dimensional spaces remains an open problem, often tackled by assuming an additive structure for f. By doing so, BO algorithms typically introduce additional restrictive assumptions on the additive structure that reduce their applicability domain. This paper contains two main contributions: (i) we relax the restrictive assumptions on the additive structure of f \textitwithout weakening the maximization guarantees of the acquisition function, and (ii) we address the over-exploration problem for decentralized BO algorithms. To these ends, we propose DuMBO, an asymptotically optimal decentralized BO algorithm that achieves very competitive performance against state-of-the-art BO algorithms, especially when the additive structure of f comprises high-dimensional factors.

   @inproceedings{dumbo,
     author = {Bardou, Anthony and Thiran, Patrick and Begin, Thomas},
     booktitle = {ICLR '24: Int'l Conference on Learning Representations},
     title = {Relaxing the Additivity Constraints in Decentralized No-Regret High-Dimensional Bayesian Optimization},
     year = {2024},
   }

3. FGCS

   NS+NDT: Smart Integration of Network Simulation in Network Digital Twin, Application to IoT Networks

   Samir Si-Mohammed, Anthony Bardou, Thomas Begin, and 2 more authors

   Future Generation Computer Systems 2024

   Abs Bib PDF

   Network Digital Twin (NDT) and Network Simulation (NS) are two paradigms leveraging virtual representations of networks to help decision-making. These tools may seem similar or interchangeable and are often confused or opposed. However, they have their respective purposes, use cases and underlying concepts, which differ and are complementary. The goal of this article is to explore and clarify the specificity, the benefits and the limits of these two decision support tools, analyze how they complement each other and can be nicely combined. We argue that a smart integration of NS in NDT, named NS+NDT, can ease, accelerate and strength decisions for network design, deployment, operations, management and evolution. To study and demonstrate this claim, we focus on the domain of Internet of Things (IoT) solutions, where wireless networks are critical for connecting the physical assets to the Internet, but are complicated to configure for meeting the requirements of a specific application. We examine how NS, coupled with NDT, can contribute to support IoT architects and operators decisions throughout the life cycle of an IoT network. We analyse the different steps required to use NS in the context of NDT and examine how this helps remove NS barriers such as credibility and reliability. In particular, we show how NDT data enable to fine tune and customize the energy consumption models, making the simulation results more context-aware and insightful. Then, addressing the often-prohibitive simulation cost for exploring a large parameter space, we propose to associate surrogate modeling to NS+NDT. As surrogate models, we first introduce a simple ML (Machine Learning)-based surrogate model and illustrate this method with two IoT network configuration optimization use cases. Secondly, we propose a Bayesian optimization approach based on Gaussian Processes as surrogate model to further accelerate the (re)configuration decisions. We show how this method enables to select simulation scenarios that converge rapidly to the optimal solution, and allows the NDT to timely perform the adaptation. The contribution of this article is threefold. It provides i) the first systematic analysis of the differences and potential synergies between NDT and NS; ii) a synthetic presentation of the integration of NS and associated decision algorithms within a NDT to unlock NS accessibility and credibility throughout the life cycle of an IoT solution; iii) a proposal for a smart and cost-efficient integration of NS in NDT via surrogate modeling, for reducing evaluation and optimization cost, paving the way to NS-augmented NDT-based dynamic adaptation and real-time optimization of IoT networks.

   @article{ns-ndt,
     author = {Si-Mohammed, Samir and Bardou, Anthony and Begin, Thomas and Lassous, Isabelle Gu{\'e}rin and Vicat-Blanc, Pascale},
     journal = {Future Generation Computer Systems},
     number = {},
     title = {NS+NDT: Smart Integration of Network Simulation in Network Digital Twin, Application to IoT Networks},
     year = {2024}
   }

2023

1. PhD ThesisBest PhD Thesis

   Online Learning for the Black-Box Optimization of Wireless Networks

   Anthony Bardou

   2023

   Abs Bib PDF Slides

   Nowadays, wireless networks (WNs) are ubiquitous and have to increase the density of their deployment to meet our important connectivity needs. Paradoxically, this deteriorates their quality of service because they have to deal with a larger amount of interference. To manage this unprecedented complexity, wireless standards, such as Wi-Fi or 5G, are becoming more flexible by introducing new degrees of freedom. Unfortunately, the correct exploitation of these parameters is not trivial because it requires to quickly identify an efficient configuration in a high-dimensional space. Moreover, the wide variety of use cases of WNs makes it difficult to model their behaviors, and consequently to find an optimal configuration analytically. In this thesis, we propose bandit and Bayesian optimization methods able to discover, through trial and error, an efficient configuration of WNs regardless of their deployments. These sequential learning methods, called "online", seek to optimize the performance of a WN by considering it as a black box. In this thesis, we illustrate the capabilities of the proposed methods on the challenging problem of spatial reuse optimization in Wi-Fi networks, and on the power control in 5G networks. Finally, we propose a new asymptotically optimal Bayesian optimization algorithm, able to optimize a high-dimensional black box function in a decentralized fashion. This last contribution could allow the implementation of more efficient protocols in WNs but also in other technological contexts.

   @phdthesis{phd,
     title = {{Online Learning for the Black-Box Optimization of Wireless Networks}},
     author = {Bardou, Anthony},
     url = {https://theses.hal.science/tel-04212534},
     number = {2023ENSL0045},
     school = {{Ecole normale sup{\'e}rieure de lyon - ENS LYON}},
     year = {2023},
     type = {Theses},
     hal_id = {tel-04212534},
     hal_version = {v2},
     bphd = {}
   }

2. COMCOM

   Analysis of a decentralized Bayesian optimization algorithm for improving spatial reuse in dense WLANs

   Anthony Bardou, and Thomas Begin

   Computer Communications 2023

   Abs Bib PDF

   Despite representing the prominent means of accessing the Internet, WLANs remain subject to performance issues, which may be mitigated through more efficient spatial reuse of radio channels. In this perspective, the IEEE 802.11ax amendment enables the dynamic update of two key parameters in wireless transmission: the transmission power (TX_POWER) and the sensitivity threshold (OBSS_PD). In this paper, we present INSPIRE, a distributed online learning solution performing local Bayesian optimizations based on Gaussian processes to improve spatial reuse in WLANs. INSPIRE makes no explicit assumptions on the WLANs’ topology and favors altruistic behaviors of the access points in their search for adequate configurations of their TX_POWER and OBSS_PD parameters. INSPIRE can easily be extended to work with a limited number of observations to throttle its computational complexity. We demonstrate the superiority of INSPIRE over other state-of-the-art strategies using the ns-3 simulator and two examples inspired by real-life deployments of dense WLANs. Our results show that, in only a few seconds, INSPIRE is able to drastically increase the quality of service of operational WLANs by improving their fairness and throughput. Finally, we discuss the configurations recommended by INSPIRE. We show that they comply with an 802.11ax empirical recommendation, and we correlate their values with some graph-based metrics of the WLAN topologies.

   @article{extension_inspire,
     author = {Bardou, Anthony and Begin, Thomas},
     journal = {Computer Communications},
     number = {},
     title = {Analysis of a decentralized Bayesian optimization algorithm for improving spatial reuse in dense WLANs},
     year = {2023}
   }

3. ADHOC

   Mitigating starvation in dense WLANs: A multi-armed Bandit Solution

   Anthony Bardou, Thomas Begin, and Anthony Busson

   Ad Hoc Networks 2023

   Abs Bib PDF

   With the recent 802.11ax amendment to the IEEE standard commercialized as Wi-Fi 6, WLANs have the potential to greatly improve the spatial reuse of radio channels. This resorts to the new ability for APs (Access Points) to dynamically modify their transmission power as well as the signal energy threshold beyond which they consider the radio channel to be free or busy. In general, selecting adequate values for these parameters is complex because of (i) the high dimensionality of the problem and (ii) the uncertainty of the radio environment. To overcome these difficulties, we frame this problem as a MAB (Multi-Armed Bandit) problem and propose an efficient and robust solution using Thompson sampling, an original sampling of WLAN configurations, and a tailor-made reward function. We evaluate the efficiency of our solution as well as several other ones with scenarios inspired by real-life WLANs’ deployments using the network simulator ns-3. The numerical results show the ability of our solution along with its superiority over the others at finding adequate parameterization at each AP thereby significantly improving the overall performance of WLANs.

   @article{adhoc,
     author = {Bardou, Anthony and Begin, Thomas and Busson, Anthony},
     journal = {Ad Hoc Networks},
     number = {},
     title = {Mitigating starvation in dense WLANs: A multi-armed Bandit Solution},
     volume = {138},
     year = {2023}
   }

2022

1. MSWiMBest Paper Award

   INSPIRE: Distributed Bayesian Optimization for ImproviNg SPatIal REuse in Dense WLANs

   Anthony Bardou, and Thomas Begin

   In MSWiM ’22: Int’l ACM Conference on Modeling Analysis and Simulation of Wireless and Mobile Systems 2022

   Abs Bib PDF

   WLANs, which have overtaken wired networks to become the primary means of connecting devices to the Internet, are prone to performance issues due to the scarcity of space in the radio spectrum. As a response, IEEE 802.11ax and subsequent amendments aim at increasing the spatial reuse of a radio channel by allowing the dynamic update of two key parameters in wireless transmission: the transmission power (TX_POWER) and the sensitivity threshold (OBSS_PD). In this paper, we present INSPIRE, a distributed online learning solution performing local Bayesian optimizations based on Gaussian processes to improve the spatial reuse in WLANs. INSPIRE makes no explicit assumptions about the topology of WLANs and favors altruistic behaviors of the access points, leading them to find adequate configurations of their TX_POWER and OBSS_PD parameters for the "greater good" of the WLANs. We demonstrate the superiority of INSPIRE over other state-of-the-art strategies using the ns-3 simulator and two examples inspired by real-life deployments of dense WLANs. Our results show that, in only a few seconds, INSPIRE is able to drastically increase the quality of service of operational WLANs by improving their fairness and throughput.

   @inproceedings{inspire,
     author = {Bardou, Anthony and Begin, Thomas},
     booktitle = {MSWiM '22: Int'l ACM Conference on Modeling Analysis and Simulation of Wireless and Mobile Systems},
     editor = {ACM},
     title = {INSPIRE: Distributed Bayesian Optimization for ImproviNg SPatIal REuse in Dense WLANs},
     year = {2022},
     bp = {},
   }

2. COMCOM

   Analysis of a Multi-Armed Bandit solution to improve the spatial reuse of next-generation WLANs

   Anthony Bardou, Thomas Begin, and Anthony Busson

   Computer Communications 2022

   Abs Bib PDF

   The next generation of WLANs will be, for the most part, ubiquitous in urban areas, densely deployed, and implementing the latest amendment of IEEE 802.11 standard, namely 802.11ax also known as Wi-Fi 6. Among the main purposes of 802.11ax is the improvement of the spatial reuse of radio channels by allowing the dynamical update of the sensitivity threshold and the transmission power at each node. In this regard, our contributions are twofold. First, we investigate the performance improvement resulting from a more efficient spatial reuse of radio channels with 802.11ax. Second, we introduce a centralized solution based on the Multi-Armed Bandit (MAB) framework and a sub-sampling technique to quickly discover an appropriate configuration of the sensitivity threshold and transmission power at each access point. We evaluate our solution with the network simulator ns-3 on different network topologies. The simulation results show the ability of our solution to quickly and robustly adjust these parameters of access points in order to significantly improve the behavior of WLANs.

   @article{extension_gaussian,
     author = {Bardou, Anthony and Begin, Thomas and Busson, Anthony},
     journal = {Computer Communications},
     number = {},
     title = {Analysis of a Multi-Armed Bandit solution to improve the spatial reuse of next-generation WLANs},
     volume = {193},
     pages = {279-292},
     year = {2022}
   }

2021

1. MSWiM

   Improving the Spatial Reuse in IEEE 802.11ax WLANs: A Multi-Armed Bandit Approach

   Anthony Bardou, Thomas Begin, and Anthony Busson

   In MSWiM ’21: Int’l ACM Conference on Modeling Analysis and Simulation of Wireless and Mobile Systems 2021

   Abs Bib PDF

   The latest amendment 802.11ax to the IEEE 802.11 standard, better known by its commercial name Wi-Fi 6, includes a feature that aims at improving the spatial reuse of a channel: each device can adapt its Clear Channel Assessment sensitivity threshold and its transmission power. In this paper, we use the Multi-Armed Bandit (MAB) framework to propose a centralized solution to dynamically adapt these parameters. We propose a new approach based on a Gaussian mixture to sample new network configurations, a specific reward function that prevents starvations when maximized, as well as a method based on Thompson Sampling to select the best network configuration. We evaluate our solution using the network simulator ns-3 and different topologies. Simulation results confirm the large benefits that 802.11ax may bring to spatial reuse. They also demonstrate the efficiency of our solution in finding appropriate parameter configurations that significantly improve the quality of service of the networks.

   @inproceedings{mswim21,
     author = {Bardou, Anthony and Begin, Thomas and Busson, Anthony},
     booktitle = {MSWiM '21: Int'l ACM Conference on Modeling Analysis and Simulation of Wireless and Mobile Systems},
     editor = {ACM},
     title = {Improving the Spatial Reuse in IEEE 802.11ax WLANs: A Multi-Armed Bandit Approach},
     year = {2021},
   }