Publications related to Ensemble Distillation for Robust Model Fusion in Federated Learning

Few-shot Learning for Efficient and Effective Machine Learning Model Adaptation

Machine learning (ML) enables artificial intelligent (AI) agents to learn autonomously from data obtained from their environment to perform tasks. Modern ML systems have proven to be extremely effective, reaching or even exceeding human intelligence. Altho ...

EPFL2024

Long plasma duration operation analyses with an international multi-machine (tokamaks and stellarators) database

Stefano Coda, Jeffrey Huang

Combined high-fusion performance and long-pulse operation is one of the key integration challenges for fusion energy development in magnetic devices. Addressing these challenges requires an integrated vision of physics and engineering aspects with the purp ...

IOP Publishing Ltd2024

Topics in statistical physics of high-dimensional machine learning

Hugo Chao Cui

In the past few years, Machine Learning (ML) techniques have ushered in a paradigm shift, allowing the harnessing of ever more abundant sources of data to automate complex tasks. The technical workhorse behind these important breakthroughs arguably lies in ...

EPFL2024

Robust machine learning for neuroscientific inference

Steffen Schneider

Modern neuroscience research is generating increasingly large datasets, from recording thousands of neurons over long timescales to behavioral recordings of animals spanning weeks, months, or even years. Despite a great variety in recording setups and expe ...

EPFL2024

On the Generalization of Stochastic Gradient Descent with Momentum

Volkan Cevher, Kimon Antonakopoulos

While momentum-based accelerated variants of stochastic gradient descent (SGD) are widely used when training machine learning models, there is little theoretical understanding on the generalization error of such methods. In this work, we first show that th ...

Microtome Publishing2024

Meta-learning to address diverse Earth observation problems across resolutions

Devis Tuia, Benjamin Alexander Kellenberger, Marc Conrad Russwurm

Earth scientists study a variety of problems with remote sensing data, but they most often consider them in isolation from each other, which limits information flows across disciplines. In this work, we present METEOR, a meta-learning methodology for Earth ...

Springernature2024

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

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

Experimental study and interpretative modelling of the Power Exhaust in Configurations with Multiple X-Points in TCV

Sophie Danielle Angelica Gorno

Nuclear fusion presents a promising clean energy source to mitigate future energy crises, with magnetic confinement fusion well-positioned to provide a baseload scenario to power future reactors. The unmitigated power exhaust of such reactors threatens its ...

EPFL2024

Federated learning with uncertainty-based client clustering for fleet-wide fault diagnosis

Hao Lu, Chao Hu, Olga Fink

Operators from various industries have been pushing the adoption of wireless sensing nodes for industrial monitoring, and such efforts have produced sizeable condition monitoring datasets that can be used to build diagnosis algorithms capable of warning ma ...

2024

Nonlinear simulation of plasma turbulence using a gyrokinetic moment-based approach

Antoine Cyril David Hoffmann

Plasma turbulence plays a fundamental role in determining the performances of magnetic confinement fusion devices, such as tokamaks. Advances in computer science, combined with the development of efficient physical models, have significantly improved our u ...

EPFL2024

Random matrix methods for high-dimensional machine learning models

Antoine Philippe Michel Bodin

In the rapidly evolving landscape of machine learning research, neural networks stand out with their ever-expanding number of parameters and reliance on increasingly large datasets. The financial cost and computational resources required for the training p ...

EPFL2024

Overview of fast particle experiments in the first MAST Upgrade experimental campaigns

Matteo Vallar

MAST-U is equipped with on-axis and off-axis neutral beam injectors (NBI), and these external sources of super-Alfv & eacute;nic deuterium fast-ions provide opportunities for studying a wide range of phenomena relevant to the physics of alpha-particles in ...

IOP Publishing Ltd2024

Global fluid simulations of plasma turbulence in stellarators

António João Caeiro Heitor Coelho

In order to cope with the decarbonization challenge faced by many countries, fusion is one of the few alternatives to fossil fuels for the production of electricity. Two devices invented in the middle of the previous century have emerged as the most promis ...

EPFL2024

Divertor turbulence characterisation using Gas Puff Imaging in the TCV tokamak

Curdin Tobias Wüthrich

The performance of magnetic confinement fusion devices, such as tokamaks, is strongly correlated to the phenomena that occur in the boundary region of the plasma core that faces the wall of the device. The dominant cross-field transport mechanisms from the ...

EPFL2024

Error field detection and correction studies towards ITER operation

Olivier Sauter, Leonardo Pigatto, Bernhard Sieglin

In magnetic fusion devices, error field (EF) sources, spurious magnetic field perturbations, need to be identified and corrected for safe and stable (disruption-free) tokamak operation. Within Work Package Tokamak Exploitation RT04, a series of studies hav ...

IOP Publishing Ltd2024

CloudProphet: A Machine Learning-Based Performance Prediction for Public Clouds

David Atienza Alonso, Ali Pahlevan, Marina Zapater Sancho, Luis Maria Costero Valero, Darong Huang

Computing servers have played a key role in developing and processing emerging compute-intensive applications in recent years. Consolidating multiple virtual machines (VMs) inside one server to run various applications introduces severe competence for limi ...

2024

First-principle based predictions of the effects of negative triangularity on DTT scenarios

Olivier Sauter, Gabriele Merlo, Alberto Mariani

Plasmas with negative triangularity (NT) shape have been recently shown to be able to achieve H-mode levels of confinement in L-mode, avoiding detrimental edge localised modes. Therefore, this plasma geometry is now studied as a possible viable option for ...

IOP Publishing Ltd2024

Unlabeled Principal Component Analysis and Matrix Completion

Liangzu Peng, Yunzhen Yao

We introduce robust principal component analysis from a data matrix in which the entries of its columns have been corrupted by permutations, termed Unlabeled Principal Component Analysis (UPCA). Using algebraic geometry, we establish that UPCA is a well-de ...

Microtome Publishing2024

Towards practical reinforcement learning for tokamak magnetic control

Jonas Buchli, Federico Alberto Alfredo Felici, Cristian Galperti, Brendan Tracey

Reinforcement learning (RL) has shown promising results for real-time control systems, including the domain of plasma magnetic control. However, there are still significant drawbacks compared to traditional feedback control approaches for magnetic confinem ...

Elsevier Science Sa2024

Global fluid simulation of plasma turbulence in stellarators with the GBS code

Joaquim Loizu Cisquella, Paolo Ricci, António João Caeiro Heitor Coelho

The implementation of three-dimensional magnetic fields, such as the ones of stellarators, in the GBS code (Ricci et al 2012 Plasma Phys. Control. Fusion54 124047; Giacomin et al 2022 J. Comput. Phys.464 111294) is presented, and simulation results are dis ...

IOP Publishing Ltd2024