Portail HAL Ensta
Not a member yet
11080 research outputs found
Sort by
A Fine-Grained Dynamic Partitioning Against Cache-Based Timing Attacks via Cache Locking
No full textInternational audienceCache-based timing side-channel attacks are prevalent and correspond to a security threat for both high-end and embedded processors. In this paper, we propose and implement a fine-grained dynamic partitioning countermeasure relying on a hardware-software collaboration. The proposed approach extends the RISC-V Instruction Set Architecture (ISA) with lock and unlock instructions to allow a program to explicitly lock cache lines in the data cache memory, ensuring constant-time accesses. Experimental results show that the proposed solution defeats contention-based cache side-channel attacks such as PRIME+PROBE and leads to a low area overhead (<3%), a low impact on binary code size (<0.3%) and a low impact on miss rate (<2%)
Laguerre–Gaussian laser filamentation for the control of electric discharges in air
No full textInternational audienceWe study the use of Laguerre–Gaussian (LG) femtosecond laser filament with multi GW peak power to guide electric sparks in the atmosphere. We demonstrate that an LG beam with a vortex phase or with 6 azimuthal phase steps generates a filamentation regime, where a longer and more uniform energy deposition is produced compared to a normal beam with a flat phase. Such filaments can guide electric discharges over much longer distances. This technique could significantly extend the guiding range of laser filaments for lightning control and other long-range atmospheric experiments involving filamentation
Etude comparative des performances des topologies d'antennes In-Band Full-Duplex mono-statiques et bi-statiques
No full textNational audienc
Simple few-shot method for spectrally resolving the wavefront of an ultrashort laser pulse
No full textInternational audienceWe present a novel, to the best of our knowledge, and straightforward approach for the spatio-spectral characterization of ultrashort pulses. This minimally intrusive method relies on placing a mask with specially arranged pinholes in the beam path before the focusing optic and retrieving the spectrally resolved laser wavefront from the speckle pattern produced at focus. We test the efficacy of this new method by accurately retrieving chromatic aberrations, such as pulse-front tilt (PFT), pulse-front curvature (PFC), and higher-order aberrations introduced by a spherical lens. The simplicity and scalability of this method, combined with its compatibility with single-shot operation, make it a strong complement to existing tools for high-intensity laser facilities
Multi-GeV Electron Acceleration in Wakefields Strongly Driven by Oversized Laser Spots
No full textInternational audienceExperiments were performed on laser wakefield acceleration in the highly nonlinear regime. With laser powers P<250 TW and using an initial spot size larger than the matched spot size for guiding, we were able to accelerate electrons to energies Emax>2.5 GeV, in fields exceeding 500 GV m−1, with more than 80 pC of charge at energies E>1 GeV. Three-dimensional particle-in-cell simulations show that using an oversized spot delays injection, avoiding beam loss as the wakefield undergoes length oscillation. This enables injected electrons to remain in the regions of highest accelerating fields and leads to a doubling of energy gain as compared to results from using half the focal length with the same laser
MAPL: Model Agnostic Peer-to-peer Learning
No full textOur code is available and can be accessed here: https://github.com/SayakMukherjee/MAPLEffective collaboration among heterogeneous clients in a decentralized setting is a rather unexplored avenue in the literature. To structurally address this, we introduce Model Agnostic Peer-to-peer Learning (coined as MAPL) a novel approach to simultaneously learn heterogeneous personalized models as well as a collaboration graph through peer-to-peer communication among neighboring clients. MAPL is comprised of two main modules: (i) local-level Personalized Model Learning (PML), leveraging a combination of intra- and inter-client contrastive losses; (ii) network-wide decentralized Collaborative Graph Learning (CGL) dynamically refining collaboration weights in a privacy-preserving manner based on local task similarities. Our extensive experimentation demonstrates the efficacy of MAPL and its competitive (or, in most cases, superior) performance compared to its centralized model-agnostic counterparts, without relying on any central server. Our code is available and can be accessed here: https://github.com/SayakMukherjee/MAP
SDEs WITH SINGULAR COEFFICIENTS: THE MARTINGALE PROBLEM VIEW AND THE STOCHASTIC DYNAMICS VIEW
No full textInternational audienceWe consider SDEs with (distributional) drift in negative Besov spaces and random initial condition and investigate them from two different viewpoints. In the first part we set up a martingale problem and show its well-posedness.We then prove further properties of the martingale problem, like continuity with respect to the drift and the link with the Fokker-Planck equation. We also show that the solutions are weak Dirichlet processes for which we evaluate the quadratic variation of the martingale component.In the second part we identify the dynamics of the solution of the martingale problemby describing the proper associated SDE.Under suitable assumptions we show equivalence with the solution to the martingale problem
On the Accessibility and Controllability of Statistical Linearization for Stochastic Control: Algebraic Rank Conditions and their Genericity
No full text23 pagesInternational audienceStatistical linearization has recently seen a particular surge of interest as a numerically cheap method for robust control of stochastic differential equations. Although it has already been successfully applied to control complex stochastic systems, accessibility and controllability properties of statistical linearization, which are key to make the robust control problem well-posed, have not been investigated yet. In this paper, we bridge this gap by providing sufficient conditions for the accessibility and controllability of statistical linearization. Specifically, we establish simple sufficient algebraic conditions for the accessibility and controllability of statistical linearization, which involve the rank of the Lie algebra generated by the drift only. In addition, we show these latter algebraic conditions are essentially sharp, by means of a counterexample, and that they are generic with respect to the drift and the initial condition
Embedded Deep Learning Accelerators: A Survey on Recent Advances
No full textInternational audienceThe exponential increase in generated data as well as the advances in high-performance computing has paved the way for the use of complex machine learning methods. Indeed, the availability of Graphical Processing Units (GPU) and Tensor Processing Units (TPU) have made it possible to train and prototype Deep Neural Networks (DNN) on large-scale data sets and for a variety of applications, i.e., vision, robotics, biomedical, etc. The popularity of these DNNs originates from their efficacy and state-of-the-art inference accuracy. However, this is obtained at the cost of a considerably high computational complexity. Such drawbacks rendered their implementation on limited resources, edge devices, without a major loss in inference speed and accuracy, a dire and challenging task. To this extent, it has become extremely important to design innovative architectures and dedicated accelerators to deploy these DNNs to embedded and re-configurable processors in a high-performance low complexity structure. In this study, we present a survey on recent advances in deep learning accelerators (DLA) for heterogeneous systems and Reduced Instruction Set Computer (RISC-V) processors given their open-source nature, accessibility, customizability and universality. After reading this article, the readers should have a comprehensive overview of the recent progress in this domain, cutting edge knowledge of recent embedded machine learning trends and substantial insights for future research directions and challenges
Méthode numérique garantie pour prouver la stabilité exponentielle des systèmes non-linéaire discret dans le temps
No full textInternational audienceThis paper presents a guaranteed numerical method for proving the exponential stability of an n-dimensional nonlinear discrete-time system. This method also finds positive invariant ellipsoids. It is based on Lyapunov theory, set theory, interval analysis, and guaranteed ellipsoidal propagation. The method is also computationally tractable and can be used on highdimensional systems where Lyapunov functions are difficult to find.Ce papier présente une méthode numérique garantie pour prouver la stabilité exponentielle d'un système non-linéaire de n-dimensions et discret dans le temps. Cette méthode trouver aussi des ellipsoïdes positive invariantes. Elle est basée sur la théorie de Lyapunov, la théorie des ensembles, l'analyse pas intervalles et la propagation garantie d'ellipsoïde. Cette méthode calculable peut être utilisée sur des systèmes de grande dimension où les fonctions de Lyapunov sont difficiles à trouver