Showing 1–12 of 12 results for author: Marty, J

Adversarial Attacks on the Interpretation of Neuron Activation Maximization

Authors: Geraldin Nanfack, Alexander Fulleringer, Jonathan Marty, Michael Eickenberg, Eugene Belilovsky

Abstract: The internal functional behavior of trained Deep Neural Networks is notoriously difficult to interpret. Activation-maximization approaches are one set of techniques used to interpret and analyze trained deep-learning models. These consist in finding inputs that maximally activate a given neuron or feature map. These inputs can be selected from a data set or obtained by optimization. However, inter… ▽ More The internal functional behavior of trained Deep Neural Networks is notoriously difficult to interpret. Activation-maximization approaches are one set of techniques used to interpret and analyze trained deep-learning models. These consist in finding inputs that maximally activate a given neuron or feature map. These inputs can be selected from a data set or obtained by optimization. However, interpretability methods may be subject to being deceived. In this work, we consider the concept of an adversary manipulating a model for the purpose of deceiving the interpretation. We propose an optimization framework for performing this manipulation and demonstrate a number of ways that popular activation-maximization interpretation techniques associated with CNNs can be manipulated to change the interpretations, shedding light on the reliability of these methods. △ Less

Submitted 12 June, 2023; originally announced June 2023.

LIO*: Low Level Information Flow Control in F*

Authors: Jean-Joseph Marty, Lucas Franceschino, Jean-Pierre Talpin, Niki Vazou

Abstract: We present Labeled Input Output in F* (LIO*), a verified framework that enforces information flow control (IFC) policies developed in F* and automatically extracted to C. Inspired by LIO, we encapsulated IFC policies into effects, but using F* we derived efficient, low level, and provably correct code. Concretely, runtime checks are lifted to static proof obligations, the developed code is automat… ▽ More We present Labeled Input Output in F* (LIO*), a verified framework that enforces information flow control (IFC) policies developed in F* and automatically extracted to C. Inspired by LIO, we encapsulated IFC policies into effects, but using F* we derived efficient, low level, and provably correct code. Concretely, runtime checks are lifted to static proof obligations, the developed code is automatically extracted to C and proved non-interferent using metaprogramming. We benchmarked our framework on three clients and observed up to 54% speedup when IFC runtime checks are proved statically. Our framework is designed to aid development of embedded devices where both enforcement of security policies and low level efficient code is critical. △ Less

Submitted 27 April, 2020; originally announced April 2020.

Comments: Submitted to ICFP

LaRa after RISE: Expected improvement in the Mars rotation and interior models

Authors: Marie-Julie Péters, Sébastien Le Maistre, Marie Yseboodt, Jean-Charles Marty, Attilio Rivoldini, Tim Van Hoolst, Véronique Dehant

Abstract: Two years after InSight's arrival at Mars, the ExoMars 2020 mission will land on the opposite side of the red planet. Similarly to InSight, which carries the RISE (Rotation and Interior Structure Experiment) radio-science experiment, the ExoMars mission will have on board the Lander Radio-science (LaRa) experiment. The X-band transponders on RISE and LaRa, allowing for direct radio-link between th… ▽ More Two years after InSight's arrival at Mars, the ExoMars 2020 mission will land on the opposite side of the red planet. Similarly to InSight, which carries the RISE (Rotation and Interior Structure Experiment) radio-science experiment, the ExoMars mission will have on board the Lander Radio-science (LaRa) experiment. The X-band transponders on RISE and LaRa, allowing for direct radio-link between the landers and stations on Earth, are dedicated to the investigation of Mars' deep interior through the precise measurement of the planet's rotation and orientation. The benefit of having LaRa after RISE for the determination of the Mars orientation and rotation parameters is demonstrated and the resulting improved constraints on the interior structure of Mars and, in particular, on its core are quantified via numerical simulations. In particular, we show that the amplitudes of the semi-annual prograde ($p_2)$ and the ter-annual retrograde ($r_3$) nutations will be determined with a precision of 6 and 4 milliarcseconds respectively by combining 700 days of RISE data with 700 days of LaRa data, about 35$\%$ more precise than what is expected from RISE alone. The impact of such an improvement on the determination of the core size of Mars is discussed and shown to be significant. △ Less

Submitted 23 October, 2019; originally announced October 2019.

The radioscience LaRa instrument onboard ExoMars 2020 to investigate the rotation and interior of Mars

Authors: Veronique Dehant, Sebastien Le Maistre, Rose-Marie Baland, Nicolas Bergeot, Ozgur Karatekin, Marie-Julie Peters, Attilio Rivoldini, Luca Ruiz Lozano, Orkun Temel, Tim Van Hoolst, Marie Yseboodt, Michel Mitrovic, Alexander Kosov, Vaclav Valenta, Lieven Thomassen, Sumit Karki, Khaldoun Al Khalifeh, Christophe Craeye, Leonid Gurvits, Jean-Charles Marty, Sami Asmar, William Folkner, the LaRa Team

Abstract: LaRa (Lander Radioscience) is an experiment on the ExoMars 2020 mission that uses the Doppler shift on the radio link due to the motion of the ExoMars platform tied to the surface of Mars with respect to the Earth ground stations (e.g. the deep space network stations of NASA), in order to precisely measure the relative velocity of the lander on Mars with respect to the Earth. The LaRa measurements… ▽ More LaRa (Lander Radioscience) is an experiment on the ExoMars 2020 mission that uses the Doppler shift on the radio link due to the motion of the ExoMars platform tied to the surface of Mars with respect to the Earth ground stations (e.g. the deep space network stations of NASA), in order to precisely measure the relative velocity of the lander on Mars with respect to the Earth. The LaRa measurements shall improve the understanding of the structure and processes in the deep interior of Mars by obtaining the rotation and orientation of Mars with a better precision compared to the previous missions. In this paper, we provide the analysis done until now for the best realization of these objectives. We explain the geophysical observation that will be reached with LaRa (Length-of-day variations, precession, nutation, and possibly polar motion). We develop the experiment set up, which includes the ground stations on Earth (so-called ground segment). We describe the instrument, i.e. the transponder and its three antennas. We further detail the link budget and the expected noise level that will be reached. Finally, we detail the expected results, which encompasses the explanation of how we shall determine Mars' orientation parameters, and the way we shall deduce Mars' interior structure and Mars' atmosphere from them. Lastly, we explain briefly how we will be able to determine the Surface platform position. △ Less

Submitted 10 October, 2019; v1 submitted 9 October, 2019; originally announced October 2019.

Comments: 43 pages, 39 figures, accepted in Planetary and Space Science

How to promote informal learning in the workplace? The need for incremental design methods

Authors: Carine Touré, Christine Michel, Jean-Charles Marty

Abstract: Informal Learning in the Workplace (ILW) is ensured by the everyday work activities in which workers are engaged. It accounts for over 75 per cent of learning in the workplace. Enterprise Social Media (ESM) are increasingly used as informal learning environments. According to the results of an implementation we have conducted in real context, we show that ESM are appropriate to promote ILW. Nevert… ▽ More Informal Learning in the Workplace (ILW) is ensured by the everyday work activities in which workers are engaged. It accounts for over 75 per cent of learning in the workplace. Enterprise Social Media (ESM) are increasingly used as informal learning environments. According to the results of an implementation we have conducted in real context, we show that ESM are appropriate to promote ILW. Nevertheless, social aspects must be reconsidered to address users' needs regarding content and access, quality information indicators, moderation and control. △ Less

Submitted 28 September, 2017; originally announced September 2017.

Journal ref: 9th International Conference on Computer Supported Education (CSEDU 2017), Feb 2017, Porto, Portugal. pp.220-229, 2017

Adaptation des r{é}seaux sociaux d'entreprise pour favoriser l'apprentissage informel sur le lieu de travail

Authors: Carine Touré, Christine Michel, Jean-Charles Marty

Abstract: The informal learning in the workplace is realized during daily collaborators' activities and represent more than 75 % of the learning occurring in a company. Enterprise social networks are currently massively used to promote this type of learning. From a pilot study, we show that they are actually adapted concerning the social aspects, but that the design must be rethought to consider user contex… ▽ More The informal learning in the workplace is realized during daily collaborators' activities and represent more than 75 % of the learning occurring in a company. Enterprise social networks are currently massively used to promote this type of learning. From a pilot study, we show that they are actually adapted concerning the social aspects, but that the design must be rethought to consider user contextual needs linked to the content and access to informational corpus, the information quality indicators and the forms of moderation and control. △ Less

Submitted 27 September, 2017; originally announced September 2017.

Comments: in French. Conf{é}rence EIAH 2017

Journal ref: 8ème Conférence sur les Environnements Informatiques pour l'Apprentissage Humain (EIAH 2017), Jun 2017, Strasbourg, France. pp.89-100, 2017

Impact analysis of the transponder time delay on radio-tracking observables

Authors: Stefano Bertone, Christophe Le Poncin-Lafitte, Pascal Rosenblatt, Valéry Lainey, Jean-Charles Marty, Marie-Christine Angonin

Abstract: Accurate tracking of probes is one of the key points of space exploration. Range and Doppler techniques are the most commonly used. In this paper we analyze the impact of the transponder delay, $i.e.$ the processing time between reception and re-emission of a two-way tracking link at the satellite, on tracking observables and on spacecraft orbits. We show that this term, only partially accounted f… ▽ More Accurate tracking of probes is one of the key points of space exploration. Range and Doppler techniques are the most commonly used. In this paper we analyze the impact of the transponder delay, $i.e.$ the processing time between reception and re-emission of a two-way tracking link at the satellite, on tracking observables and on spacecraft orbits. We show that this term, only partially accounted for in the standard formulation of computed space observables, can actually be relevant for future missions with high nominal tracking accuracies or for the re-processing of old missions. We present several applications of our formulation to Earth flybys, the NASA GRAIL and the ESA BepiColombo missions. △ Less

Submitted 5 September, 2017; v1 submitted 1 August, 2017; originally announced August 2017.

Comments: 19 pages, 4 figures, 2 tables. Accepted for publication in Advances in Space Research

Planetary Radio Interferometry and Doppler Experiment (PRIDE) technique: A test case of the Mars Express Phobos fly-by

Authors: Dmitry A. Duev, Sergei V. Pogrebenko, Giuseppe Cimò, Guifré Molera Calvés, Tatiana M. Bocanegra Bahamón, Leonid I. Gurvits, Mark M. Kettenis, Joseph Kania, Valeriu Tudose, Pascal Rosenblatt, Jean-Charles Marty, Valery Lainey, Pablo de Vicente, Jonathan Quick, Marisa Nickola, Alexander Neidhardt, Gerhard Kronschnabl, Christian Plötz, Rüdiger Haas, Michael Lindqvist, Andrea Orlati, Alexander V. Ipatov, Mikhail A. Kharinov, Andrey G. Mikhailov, Jim Lovell , et al. (11 additional authors not shown)

Abstract: The closest ever fly-by of the Martian moon Phobos, performed by the European Space Agency's Mars Express spacecraft, gives a unique opportunity to sharpen and test the Planetary Radio Interferometry and Doppler Experiments (PRIDE) technique in the interest of studying planet - satellite systems. The aim of this work is to demonstrate a technique of providing high precision positional and Doppler… ▽ More The closest ever fly-by of the Martian moon Phobos, performed by the European Space Agency's Mars Express spacecraft, gives a unique opportunity to sharpen and test the Planetary Radio Interferometry and Doppler Experiments (PRIDE) technique in the interest of studying planet - satellite systems. The aim of this work is to demonstrate a technique of providing high precision positional and Doppler measurements of planetary spacecraft using the Mars Express spacecraft. The technique will be used in the framework of Planetary Radio Interferometry and Doppler Experiments in various planetary missions, in particular in fly-by mode. We advanced a novel approach to spacecraft data processing using the techniques of Doppler and phase-referenced very long baseline interferometry spacecraft tracking. We achieved, on average, mHz precision (30 μm/s at a 10 seconds integration time) for radial three-way Doppler estimates and sub-nanoradian precision for lateral position measurements, which in a linear measure (at a distance of 1.4 AU) corresponds to ~50 m. △ Less

Submitted 19 June, 2016; originally announced June 2016.

Comments: 9 pages, 14 figures. Astronomy and Astrophysics, accepted on 2016/05/31

Journal ref: A&A 593, A34 (2016)

Re-designing knowledge management systems : Towards user-centred design methods integrating information architecture

Authors: Carine Touré, Christine Michel, Jean-Charles Marty

Abstract: The work presented in this paper focuses on the improvement of corporate knowledge management systems. For the implementation of such systems, companies deploy can important means for small gains. Indeed, management services often notice very limited use compared to what they actually expect. We present a five-step redesigning approach which takes into account different factors to increase the use… ▽ More The work presented in this paper focuses on the improvement of corporate knowledge management systems. For the implementation of such systems, companies deploy can important means for small gains. Indeed, management services often notice very limited use compared to what they actually expect. We present a five-step redesigning approach which takes into account different factors to increase the use of these systems. We use as an example the knowledge sharing platform implemented for the employees of Soci{é}t{é} du Canal de Provence (SCP). This system was taken into production but very occasionally used. We describe the reasons for this limited use and we propose a design methodology adapted to the context. Promoting the effective use of the system, our approach has been experimented and evaluated with a panel of users working at SCP. △ Less

Submitted 29 January, 2016; originally announced January 2016.

Comments: in Knowledge Management and Information Sharing, Oct 2014, Rome, Italy

What if we considered awareness for sustainable Kwowledge Management ? Towards a model for self regulated knowledge management systems based on acceptance models of technologies and awareness

Authors: Carine Touré, Christine Michel, Jean-Charles Marty

Abstract: We propose, in this paper, a model of continuous use of corporate collaborative KMS. Companies do not always have the guaranty that their KMS will be continuously used. This statement can constitute an important obstacle for knowledge management processes. Our work is based on the analysis of classical models for initial and continuous use of technologies. We also analyse the regulation concept an… ▽ More We propose, in this paper, a model of continuous use of corporate collaborative KMS. Companies do not always have the guaranty that their KMS will be continuously used. This statement can constitute an important obstacle for knowledge management processes. Our work is based on the analysis of classical models for initial and continuous use of technologies. We also analyse the regulation concept and explain how it is valuable to support a continuous use of KMS. We observed that awareness may be a regulation means that allows taking this problem into account. Awareness is a concept, which has been profusely used to improve user experience in collaborative environments. It is an important element for regulation of activity. In our model, we assume that one can integrate awareness in information systems to positively influence beliefs about them. The final objective of our work is to refine some concepts to fit the particularities of collaborative KMS and to propose an awareness regulation process using the traces of the users' interactions with the systems. △ Less

Submitted 29 January, 2016; originally announced January 2016.

Comments: Knowledge Management and Information Sharing (KMIS), Oct 2014, Rome, Italy. 2014

Trans-gram, Fast Cross-lingual Word-embeddings

Authors: Jocelyn Coulmance, Jean-Marc Marty, Guillaume Wenzek, Amine Benhalloum

Abstract: We introduce Trans-gram, a simple and computationally-efficient method to simultaneously learn and align wordembeddings for a variety of languages, using only monolingual data and a smaller set of sentence-aligned data. We use our new method to compute aligned wordembeddings for twenty-one languages using English as a pivot language. We show that some linguistic features are aligned across languag… ▽ More We introduce Trans-gram, a simple and computationally-efficient method to simultaneously learn and align wordembeddings for a variety of languages, using only monolingual data and a smaller set of sentence-aligned data. We use our new method to compute aligned wordembeddings for twenty-one languages using English as a pivot language. We show that some linguistic features are aligned across languages for which we do not have aligned data, even though those properties do not exist in the pivot language. We also achieve state of the art results on standard cross-lingual text classification and word translation tasks. △ Less

Submitted 11 January, 2016; originally announced January 2016.

Comments: EMNLP 2015

Effects of transition metal substitutions on the incommensurability and spin fluctuations in BaFe2As2 by elastic and inelastic neutron scattering

Authors: M. G. Kim, J. Lamsal, T. W. Heitmann, G. S. Tucker, D. K. Pratt, S. N. Khan, Y. B. Lee, A. Alam, A. Thaler, N. Ni, S. Ran, S. L. Bud'ko, K. J. Marty, M. D. Lumsden, P. C. Canfield, B. N. Harmon, D. D. Johnson, A. Kreyssig, R. J. McQueeney, A. I. Goldman

Abstract: The spin fluctuation spectra from nonsuperconducting Cu-substituted, and superconducting Co-substituted, BaFe2As2 are compared quantitatively by inelastic neutron scattering measurements and are found to be indis- tinguishable. Whereas diffraction studies show the appearance of incommensurate spin-density wave order in Co and Ni substituted samples, the magnetic phase diagram for Cu substitution d… ▽ More The spin fluctuation spectra from nonsuperconducting Cu-substituted, and superconducting Co-substituted, BaFe2As2 are compared quantitatively by inelastic neutron scattering measurements and are found to be indis- tinguishable. Whereas diffraction studies show the appearance of incommensurate spin-density wave order in Co and Ni substituted samples, the magnetic phase diagram for Cu substitution does not display incommensu- rate order, demonstrating that simple electron counting based on rigid-band concepts is invalid. These results, supported by theoretical calculations, suggest that substitutional impurity effects in the Fe plane play a signifi- cant role in controlling magnetism and the appearance of superconductivity, with Cu distinguished by enhanced impurity scattering and split-band behavior. △ Less

Submitted 19 October, 2012; v1 submitted 6 April, 2012; originally announced April 2012.

Comments: 5 pages, 5 figures, Major change in the manuscript

Journal ref: Phys. Rev. Lett. 109, 167003 (2012)