Publications associées à Adresse électronique

Byzantine Machine Learning Made Easy By Resilient Averaging of Momentums

Rachid Guerraoui, John Stephan, Sadegh Farhadkhani, Nirupam Gupta, Rafaël Benjamin Pinot

Byzantine resilience emerged as a prominent topic within the distributed machine learning community. Essentially, the goal is to enhance distributed optimization algorithms, such as distributed SGD, in a way that guarantees convergence despite the presence ...

PMLR2022

Pretraining boosts out-of-domain robustness for pose estimation

Alexander Mathis, Mackenzie Mathis, Steffen Schneider, Matthias Bethge, Thomas Ray Biasi, Mert Yüksekgönül

Neural networks are highly effective tools for pose estimation. However, as in other computer vision tasks, robustness to out-of-domain data remains a challenge, especially for small training sets that are common for real-world applications. Here, we probe ...

IEEE COMPUTER SOC2021

GARFIELD: System Support for Byzantine Machine Learning (Regular Paper)

Rachid Guerraoui, Sébastien Louis Alexandre Rouault, Arsany Hany Abdelmessih Guirguis

We present GARFIELD, a library to transparently make machine learning (ML) applications, initially built with popular (but fragile) frameworks, e.g., TensorFlow and PyTorch, Byzantine–resilient. GARFIELD relies on a novel object–oriented design, reducing t ...

IEEE2021

Template Induction over Unstructured Email Corpora

Julia Proskurnia

Unsupervised template induction over email data is a central component in applications such as information extraction, document classification, and auto-reply. The benefits of automatically generating such templates are known for structured data, e.g. mach ...

Profiling, Modelling and Facilitating Online Activism

Julia Proskurnia

The extensive and successful use of social media has enhanced and empowered a variety of movements all over the world in a way that is hard to achieve through conventional means. This has led to numerous studies that leverage online social data to describe ...

EPFL2017

Reconstruction of 3D scattered data via radial basis functions by efficient and robust techniques

Alberto Crivellaro

We propose new algorithms to overcome two of the most constraining limitations of surface reconstruction methods in use. In particular, we focus on the large amount of data characterizing standard acquisitions by scanner and the noise intrinsically introdu ...

Elsevier Science Bv2017

JammJoint: A Variable Stiffness Device Based on Granular Jamming for Wearable Joint Support

Auke Ijspeert, Jamie Paik, Simon Lukas Hauser, Matthew Aaron Robertson

In robotics, controlling the stiffness of the joints that contribute to the robots’ degree of freedom dictates the adaptability, versatility, and safety of the whole system.We can achieve variable stiffness or impedance in a robotic system purely by the co ...

2017

Towards a Principled Approach to Humanitarian Information and Communication Technology

Denis Gillet, Isabelle Barbara Marie-Hélène Cardia, Adrian Christian Holzer

The humanitarian sector is entering in the network age. The explosion in access to new information and communication technology (ICT) like mobile phones, the internet, and social media, is changing the way people communicate in emergency situations. Howeve ...

ACM2017

Remotely Controlled Isomer Selective Molecular Switching

Klaus Kern, Uta Schlickum, Hagen Klauk, Peter Wahl, Ivan Pentegov, Verena Katharina Schendel

Non local addressing-the "remote control"- of molecular switches promises more efficient processing for information technology, where fast speed of switching is essential. The surface state of the (111) facets of noble metals, a confined two-dimensional el ...

Amer Chemical Soc2016

Admissible geometrical domains and graphic statics to evaluate constitutive elements of structural robustness

Corentin Jean Dominique Fivet

Structural robustness is related to the insensitivity of a structure to local failure, which is linked to the ability of force redistribution. Current approaches—probabilistic, risk-based, deterministic, and energetic—are poorly suited to inform the design ...

2016