Publications associées à Quadripôle

Deep Learning Theory Through the Lens of Diagonal Linear Networks

In this PhD manuscript, we explore optimisation phenomena which occur in complex neural networks through the lens of

2

-layer diagonal linear networks. This rudimentary architecture, which consists of a two layer feedforward linear network with a diagonal ...

EPFL2024

Pyroresistive response of percolating conductive polymer composites

Claudio Grimaldi

The pyroresistive response of conductive polymer composites (CPCs) has attracted much interest because of its potential applications in many electronic devices requiring a significant responsiveness to changes in external physical parameters such as temper ...

Amer Physical Soc2024

Unified Harmonic Power-Flow and Stability Analysis of Power Grids with Converter-Interfaced Distributed Energy Resources

Johanna Kristin Maria Becker

Modern power distribution systems are experiencing a large-scale integration of Converter-Interfaced Distributed Energy Resources (CIDERs). Their presence complicates the analysis and mitigation of harmonics, whose creation and propagation may be amplified ...

EPFL2024

Large reconfigurable quantum circuits with SPAD arrays and multimode fibers

Edoardo Charbon, Claudio Bruschini, Ivan Michel Antolovic

Reprogrammable integrated optics provides a natural platform for tunable quantum photonic circuits, but faces challenges when high dimensions and high connectivity are involved. Here, we implement high-dimensional linear transformations on spatial modes of ...

Optica Publishing Group2024

The effect of plasma shaping on high density H-mode SOL profiles and fluctuations in TCV

Holger Reimerdes, Benoît Labit, Christian Gabriel Theiler, Umar Sheikh, Guang-Yu Sun, Luke Simons, Nicola Vianello, Sophie Danielle Angelica Gorno, Cedric Kar-Wai Tsui, Claudia Colandrea, Adriano Stagni, Yi Wang

The impact of plasma shaping on the properties of high density H-mode scrape-off layer (SOL) profiles and transport at the outer midplane has been investigated on Tokamaka configuration variable. The experimental dataset has been acquired by evolving the u ...

IOP Publishing Ltd2024

Ripple Voltage and Loss Reduction of Single-Phase ISOP SST by Eliminating the Second Harmonic Current in LLC Converters

Drazen Dujic, Andrea Cervone, Tianyu Wei

In two-stage solid-state transformers with input-series output-parallel structure, the LLC converter operating at a fixed switching frequency is a common choice for the isolated DC/DC conversion stage because it can provide a high efficiency and a fixed vo ...

2024

Role of Minimum Adhesive Wear Particle Size in Third-Body Layer Properties

Jean-François Molinari, Son-Jonathan Pham-Ba

We employ a novel discrete element method (DEM) force formulation to simulate adhesive wear and assess the effects of material and loading parameters on the properties of the third-body layer (TBL) formed during sliding motion. The study emphasizes the rol ...

2023

Design and Development of a Diagnostic System for a Non-Intercepting Direct Measure of the SPIDER Ion Source Beamlet Current

Stable and uniform beams with low divergence are required in particle accelerators; therefore, beyond the accelerated current, measuring the beam current spatial uniformity and stability over time is necessary to assess the beam performance, since these pa ...

MDPI2023

Saddle-to-Saddle Dynamics in Diagonal Linear Networks

Nicolas Henri Bernard Flammarion, Scott William Pesme

In this paper we fully describe the trajectory of gradient flow over diagonal linear networks in the limit of vanishing initialisation. We show that the limiting flow successively jumps from a saddle of the training loss to another until reaching the minim ...

2023

Incremental Learning in Diagonal Linear Networks

Raphaël Jean Berthier

Diagonal linear networks (DLNs) are a toy simplification of artificial neural networks; they consist in a quadratic reparametrization of linear regression inducing a sparse implicit regularization. In this paper, we describe the trajectory of the gradient ...

Microtome Publishing2023

Beam homogeneity of caesium seeded SPIDER using a direct beamlet current measurement

The ITER Heating Neutral Beam (HNB) source prototype SPIDER (Source for the Production of Ions of Deuterium Extracted from a Radio frequency plasma), hosted at the Neutral Beam Test Facility (NBTF) in Padova, Italy, has recently started operating with evap ...

ELSEVIER SCIENCE SA2023

(S)GD over Diagonal Linear Networks: Implicit Regularisation, Large Stepsizes and Edge of Stability

Nicolas Henri Bernard Flammarion, Scott William Pesme, Mathieu Even

In this paper, we investigate the impact of stochasticity and large stepsizes on the implicit regularisation of gradient descent (GD) and stochastic gradient descent (SGD) over diagonal linear networks. We prove the convergence of GD and SGD with macroscop ...

2023

Performance of the Cable-in-Conduit Conductors for Super-X Test Facility

Kamil Sedlák

Following the conceptual design, the engineering design of the dc magnet and cable-in-conduit conductor (CICC) for the Super-X test facility has been done in 2021. Totally three types of conductors with different structures were designed for the three pair ...

IEEE Institute of Electrical and Electronics Engineers2023

Detecting Anomalies and Obstacles in Road Scenes

Krzysztof Maciej Lis

We address the problem of segmenting anomalies and unusual obstacles in road scenes for the purpose of self-driving safety. The objects in question are not present in the common training sets as it is not feasible to collect and annotate examples for every ...

EPFL2023

Latent Space Slicing for Enhanced Entropy Modeling in Learning-Based Point Cloud Geometry Compression

Touradj Ebrahimi, Davi Nachtigall Lazzarotto

The growing adoption of point clouds as an imaging modality has stimulated the search for efficient solutions for compression. Learning-based algorithms have been reporting increasingly better performance and are drawing the attention from the research com ...

IEEE2022

Role of stochastic noise and generalization error in the time propagation of neural-network quantum states

Giuseppe Carleo

Neural-network quantum states (NQS) have been shown to be a suitable variational ansatz to simulate out-of-equilibrium dynamics in two-dimensional systems using timedependent variational Monte Carlo (t-VMC). In particular, stable and accurate time propagat ...

SCIPOST FOUNDATION2022

Evaluation Criterion of Proton Exchange Membrane (ECPEM) fuel cells considering inserted porous media inside the gas flow channel

Jan Van Herle, Hossein Pourrahmani

The focus of this study is to evaluate the effects of porous media inside the gas flow channel of Proton Exchange Membrane Fuel Cells (PEMFC) on four different output parameters of voltage, power density, pressure drop, and Nusselt number (Nu) considering ...

PERGAMON-ELSEVIER SCIENCE LTD2022

EMC Impact of Disturbances Generated by Multiple Sources

Farhad Rachidi-Haeri, Marcos Rubinstein, Hamidreza Karami

In this paper, the impact of an increasing number of arbitrary electrical/electronic devices on the overall radiated emissions is investigated. Understanding and quantifying such an impact are prerequisites to the proper evaluation of electromagnetic compa ...

2022

Towards Understanding Sharpness-Aware Minimization

Nicolas Henri Bernard Flammarion, Maksym Andriushchenko

Sharpness-Aware Minimization (SAM) is a recent training method that relies on worst-case weight perturbations which significantly improves generalization in various settings. We argue that the existing justifications for the success of SAM which are based ...

JMLR-JOURNAL MACHINE LEARNING RESEARCH2022

Step Change Detection for Improved ROCOF Evaluation of Power System Waveforms

Mario Paolone, Asja Derviskadic, Alexandra Cameron Karpilow, Guglielmo Frigo

In the analysis of power grid waveforms, the presence of amplitude or phase steps can disrupt the estimation of frequency and rate-of-change-of-frequency (ROCOF). Standard methods based on phasor-models fail in the extraction of signal parameters during th ...

IEEE2022