Publications

2017

• Dynamic mitigation of EDFA power excursions with machine learning
  
  • Huang Yishen
  • Gutterman Craig L.
  • Samadi Payman
  • Cho Patricia B.
  • Samoud Wiem
  • Ware Cédric
  • Lourdiane Mounia
  • Zussman Gil
  • Bergman Keren
  Optics Express, Optical Society of America - OSA Publishing, 2017, 25 (3), pp.2245 - 2258. Dynamic optical networking has promising potential to support the rapidly changing traffic demands in metro and long-haul networks. However, the improvement in dynamicity is hindered by wavelength-dependent power excursions in gain-controlled erbium doped fiber amplifiers (EDFA) when channels change rapidly. We introduce a general approach that leverages machine learning (ML) to characterize and mitigate the power excursions of EDFA systems with different equipment and scales. An ML engine is developed and experimentally validated to show accurate predictions of the power dynamics in cascaded EDFAs. Recommended channel provisioning based on the ML predictions achieves within 1% error of the lowest possible power excursion over 94% of the time. We also showcase significant mitigation of EDFA power excursions in super-channel provisioning when compared to the first-fit wavelength assignment algorithm (10.1364/OE.25.002245)
  DOI : 10.1364/OE.25.002245
• Décomposition de séries temporelles d'images SAR pour la détection de changements
  
  • Lobry Sylvain
  • Denis L.
  • Zhao Weiying
  • Tupin Florence
  Traitement du Signal, Lavoisier, 2017.
• Robust Downbeat Tracking Using an Ensemble of Convolutional Networks
  
  • Durand Simon
  • Bello Juan Pablo
  • David Bertrand
  • Richard Gael
  IEEE/ACM Transactions on Audio, Speech and Language Processing, Institute of Electrical and Electronics Engineers, 2017, 25 (1), pp.76-89. <div><p>In this paper, we present a novel state of the art system for automatic downbeat tracking from music signals. The audio signal is first segmented in frames which are synchronized at the tatum level of the music. We then extract different kind of features based on harmony, melody, rhythm and bass content to feed convolutional neural networks that are adapted to take advantage of the characteristics of each feature. This ensemble of neural networks is combined to obtain one downbeat likelihood per tatum. The downbeat sequence is finally decoded with a flexible and efficient temporal model which takes advantage of the assumed metrical continuity of a song. We then perform an evaluation of our system on a large base of 9 datasets, compare its performance to 4 other published algorithms and obtain a significant increase of 16.8 percent points compared to the second best system, for altogether a moderate cost in test and training. The influence of each step of the method is studied to show its strengths and shortcomings.</p></div> (10.1109/TASLP.2016.2623565)
  DOI : 10.1109/TASLP.2016.2623565
• Parameter estimation of perfusion models in dynamic contrast-enhanced imaging: a unified framework for model comparison
  
  • Romain Blandine
  • Rouet Laurence
  • Ohayon Daniel
  • Lucidarme Olivier
  • d'Alché-Buc Florence
  • Letort Véronique
  Medical Image Analysis, Elsevier, 2017, 35, pp.360--374. Patients follow-up in oncology is generally performed through the acquisition of dynamic sequences of contrast-enhanced images. Estimating parameters of appropriate models of contrast intake diffusion through tissues should help characterizing the tumour physiology. However, several models have been developed and no consensus exists on their clinical use. In this paper, we propose a unified framework to analyse models of perfusion and estimate their parameters in order to obtain reliable and relevant parametric images. After defining the biological context and the general form of perfusion models, we propose a methodological framework for model assessment in the context of parameter estimation from dynamic imaging data: global sensitivity analysis, structural and practical identifiability analysis, parameter estimation and model comparison. Then, we apply our methodology to five of the most widely used compartment models (Tofts model, extended Tofts model, two-compartment model, tissue-homogeneity model and distributed-parameters model) and illustrate the results by analysing the behaviour of these models when applied to data acquired on five patients with abdominal tumours. (10.1016/j.media.2016.07.008)
  DOI : 10.1016/j.media.2016.07.008
• Déborder l'expérience pour laisser une trace. Vidéophénoménographie d'un rappeur
  
  • Kneubühler Marine
  , 2017, pp.147-242.
• On some bounds for symmetric tensor rank of multiplication in finite fields
  
  • Ballet Stéphane
  • Pieltant Julia
  • Rambaud Matthieu
  • Sijsling Jeroen
  , 2017, 686, pp.93 - 121. We establish new upper bounds about symmetric bilinear complexity in any extension of finite fields. Note that these bounds are not asymptotical but uniform. Moreover, we discuss the validity of certain published bounds. (10.1090/conm/686/13779)
  DOI : 10.1090/conm/686/13779
• Conception d'absorbants à base de métasurfaces et Application de la Transformation d'Espace pour le contrôle du rayonnement d’une antenne imprimée
  
  • Lepage A. C.
  • Begaud Xavier
  , 2017.
• LISP EID Block
  
  • Iannone Luigi
  • Lewis Darrel
  • Meyer Dave
  • Fuller Vince
  , 2017.
• On perturbed proximal gradient algorithms
  
  • Atchadé Yves
  • Fort Gersende
  • Moulines Éric
  Journal of Machine Learning Research, Microtome Publishing, 2017, 18 (10), pp.1-33.
• Demystifying the asymptotic behavior of global denoising
  
  • Houdard Antoine
  • Almansa Andrès
  • Delon Julie
  Journal of Mathematical Imaging and Vision, Springer Verlag, 2017. In this work, we revisit the global denoising framework recently introduced by Talebi and Milanfar. We analyze the asymptotic behavior of its mean-squared error restoration performance in the oracle case when the image size tends to infinity. We introduce precise conditions on both the image and the global filter to ensure and quantify this convergence. We also make a clear distinction between two different levels of oracle that are used in that framework. By reformulating global denoising with the classical formalism of diagonal estimation, we conclude that the second-level oracle can be avoided by using Donoho and Johnstone’s theorem, whereas the first-level oracle is mostly required in the sequel. We also discuss open issues concerning the most challenging aspect, namely the extension of these results to the case where neither oracle is required. (10.1007/s10851-017-0716-6)
  DOI : 10.1007/s10851-017-0716-6
• Planck intermediate results. LI. Features in the cosmic microwave background temperature power spectrum and shifts in cosmological parameters
  
  • Aghanim N.
  • Akrami Yashar
  • Ashdown M.
  • Aumont J.
  • Ballardini M.
  • Banday A.J.
  • Barreiro R.B.
  • Bartolo N.
  • Basak S.
  • Benabed K.
  • Bersanelli M.
  • Bielewicz P.
  • Bonaldi A.
  • Bonavera L.
  • Bond J.R.
  • Borrill J.
  • Bouchet F.R.
  • Burigana C.
  • Calabrese E.
  • Cardoso J.F.
  • Challinor A.
  • Chiang H.C.
  • Colombo L.P.L.
  • Combet C.
  • Crill B.P.
  • Curto A.
  • Cuttaia F.
  • de Bernardis P.
  • de Rosa A.
  • de Zotti G.
  • Delabrouille J.
  • Di Valentino E.
  • Dickinson C.
  • Diego J.M.
  • Dore O.
  • Ducout A.
  • Dupac X.
  • Dusini S.
  • Efstathiou G.
  • Elsner F.
  • Ensslin T.A.
  • Eriksen H.K.
  • Fantaye Y.
  • Finelli F.
  • Forastieri F.
  • Frailis M.
  • Franceschi E.
  • Frolov A.
  • Galeotta S.
  • Galli S.
  • Ganga K.
  • Genova-Santos R.T.
  • Gerbino M.
  • Gonzalez-Nuevo J.
  • Gorski K.M.
  • Gruppuso A.
  • Gudmundsson J.E.
  • Herranz D.
  • Hivon E.
  • Huang Z.
  • Jaffe A.H.
  • Jones W.C.
  • Keihanen E.
  • Keskitalo R.
  • Kiiveri K.
  • Kim J.
  • Kisner T.S.
  • Knox L.
  • Krachmalnicoff N.
  • Kunz M.
  • Kurki-Suonio H.
  • Lagache G.
  • Lamarre J.M.
  • Lasenby A.
  • Lattanzi M.
  • Lawrence C.R.
  • Le Jeune M.
  • Levrier F.
  • Lewis A.
  • Lilje P.B.
  • Lilley M.
  • Lindholm V.
  • Lopez-Caniego M.
  • Lubin P.M.
  • Ma Y.Z.
  • Macias-Perez J.F.
  • Maggio G.
  • Maino D.
  • Mandolesi N.
  • Mangilli A.
  • Maris M.
  • Martin P.G.
  • Martinez-Gonzalez E.
  • Matarrese S.
  • Mauri N.
  • Mcewen J.D.
  • Meinhold P.R.
  • Mennella A.
  • Migliaccio M.
  • Millea M.
  • Miville-Deschenes M.A.
  • Molinari D.
  • Moneti A.
  • Montier L.
  • Morgante G.
  • Moss A.
  • Narimani A.
  • Natoli P.
  • Oxborrow C.A.
  • Pagano L.
  • Paoletti D.
  • Patanchon G.
  • Patrizii L.
  • Pettorino V.
  • Piacentini F.
  • Polastri L.
  • Polenta G.
  • Puget J.L.
  • Rachen J.P.
  • Racine B.
  • Reinecke M.
  • Remazeilles M.
  • Renzi A.
  • Rossetti M.
  • Roudier G.
  • Rubino-Martin J.A.
  • Ruiz-Granados B.
  • Salvati L.
  • Sandri M.
  • Savelainen M.
  • Scott D.
  • Sirignano C.
  • Sirri G.
  • Stanco L.
  • Suur-Uski A.S.
  • Tauber J.A.
  • Tavagnacco D.
  • Tenti M.
  • Toffolatti L.
  • Tomasi M.
  • Tristram M.
  • Trombetti T.
  • Valiviita J.
  • van Tent F.
  • Vielva P.
  • Villa Francesca
  • Vittorio N.
  • Wandelt B.D.
  • Wehus I.K.
  • White M.
  • Zacchei A.
  • Zonca A.
  Astronomy & Astrophysics - A&A, EDP Sciences, 2017, 607, pp.A95. The six parameters of the standard ΛCDM model have best-fit values derived from the Planck temperature power spectrum that are shifted somewhat from the best-fit values derived from WMAP data. These shifts are driven by features in the Planck temperature power spectrum at angular scales that had never before been measured to cosmic-variance level precision. We have investigated these shifts to determine whether they are within the range of expectation and to understand their origin in the data. Taking our parameter set to be the optical depth of the reionized intergalactic medium τ, the baryon density ωb, the matter density ωm, the angular size of the sound horizon θ∗, the spectral index of the primordial power spectrum, ns, and Ase− 2τ (where As is the amplitude of the primordial power spectrum), we have examined the change in best-fit values between a WMAP-like large angular-scale data set (with multipole moment ℓ < 800 in the Planck temperature power spectrum) and an all angular-scale data set (ℓ < 2500Planck temperature power spectrum), each with a prior on τ of 0.07 ± 0.02. We find that the shifts, in units of the 1σ expected dispersion for each parameter, are { Δτ,ΔAse− 2τ,Δns,Δωm,Δωb,Δθ∗ } = { −1.7,−2.2,1.2,−2.0,1.1,0.9 }, with a χ2 value of 8.0. We find that this χ2 value is exceeded in 15% of our simulated data sets, and that a parameter deviates by more than 2.2σ in 9% of simulated data sets, meaning that the shifts are not unusually large. Comparing ℓ < 800 instead to ℓ> 800, or splitting at a different multipole, yields similar results. We examined the ℓ < 800 model residuals in the ℓ> 800 power spectrum data and find that the features there that drive these shifts are a set of oscillations across a broad range of angular scales. Although they partly appear similar to the effects of enhanced gravitational lensing, the shifts in ΛCDM parameters that arise in response to these features correspond to model spectrum changes that are predominantly due to non-lensing effects; the only exception is τ, which, at fixed Ase− 2τ, affects the ℓ> 800 temperature power spectrum solely through the associated change in As and the impact of that on the lensing potential power spectrum. We also ask, “what is it about the power spectrum at ℓ < 800 that leads to somewhat different best-fit parameters than come from the full ℓ range?” We find that if we discard the data at ℓ < 30, where there is a roughly 2σ downward fluctuation in power relative to the model that best fits the full ℓ range, the ℓ < 800 best-fit parameters shift significantly towards the ℓ < 2500 best-fit parameters. In contrast, including ℓ < 30, this previously noted “low-ℓ deficit” drives ns up and impacts parameters correlated with ns, such as ωm and H0. As expected, the ℓ < 30 data have a much greater impact on the ℓ < 800 best fit than on the ℓ < 2500 best fit. So although the shifts are not very significant, we find that they can be understood through the combined effects of an oscillatory-like set of high-ℓ residuals and the deficit in low-ℓ power, excursions consistent with sample variance that happen to map onto changes in cosmological parameters. Finally, we examine agreement between PlanckTT data and two other CMB data sets, namely the Planck lensing reconstruction and the TT power spectrum measured by the South Pole Telescope, again finding a lack of convincing evidence of any significant deviations in parameters, suggesting that current CMB data sets give an internally consistent picture of the ΛCDM model.Key words: cosmology: observations / cosmic background radiation / cosmological parameters / cosmology: theory (10.1051/0004-6361/201629504)
  DOI : 10.1051/0004-6361/201629504
• Nonequilibrium Green's functions theory for the alpha factor of quantum cascade lasers
  
  • Pereira Mauro
  • Winge David
  • Wacker Andreas
  • Jumpertz Louise
  • Michel Florian
  • Pawlus Robert
  • Elsassaer Wolfgang
  • Schires Kevin
  • Carras Mathieu
  • Grillot Frédéric
  , 2017.
• Étude des perturbations des systèmes de positionnement magnéto-inductifs en intérieur, » Journées nationales micorondes
  
  • Gharat Vighnesh
  • Colin Elizabeth
  • Baudoin Geneviève
  • Richard Damien Richard
  , 2017.
• Beat Gesture Prediction using Prosodic Features
  
  • Jain Varun
  • Clavel Chloé
  • Pelachaud Catherine I
  , 2017.
• Fuzzy Skeleton and Skeleton by Influence Zones: A Review
  
  • Bloch Isabelle
  , 2017, pp.71-87. <p> Skeleton is a widely addressed topic in binary image processing and object or shape representation. The difficulty raised by the discrete nature of images has led to different classes of methods. Among the most popular ones, distance based methods led to the important notion of centers of maximal balls (CMB), and mathematical morphology based methods led both to an efficient computation of CMB and to a completely different class of approaches, relying on homotopic thinning. Skeleton by influence zones (SKIZ) is another important related problem, often addressed using tools from mathematical morphology.</p> <p> </p> <p>When imprecision has to be explicitly modeled, then objects become fuzzy sets and all the previous approaches for skeleton and SKIZ have to be extended to deal with fuzzy sets and to cope with spatial imprecision. This chapter gives an overview of the main existing definitions of fuzzy skeleton and fuzzy SKIZ, including some derived from grey level image processing, and proposes a few novel definitions. </p>
• Method, device, and computer program for transmitting portions of encapsulated media content
  
  • Denoual Franck
  • Mazé Frédéric
  • Ruellan Hervé
  • Le Feuvre J.
  • Ouedraogo Nael
  , 2017.
• Autoreject: Automated artifact rejection for MEG and EEG data
  
  • Jas Mainak
  • Engemann Denis A
  • Bekhti Yousra
  • Raimondo Federico A
  • Gramfort Alexandre
  NeuroImage, Elsevier, 2017. We present an automated algorithm for unified rejection and repair of bad trials in magnetoencephalography (MEG) and electroencephalography (EEG) signals. Our method capitalizes on cross-validation in conjunction with a robust evaluation metric to estimate the optimal peak-to-peak threshold – a quantity commonly used for identifying bad trials in M/EEG. This approach is then extended to a more sophisticated algorithm which estimates this threshold for each sensor yielding trial-wise bad sensors. Depending on the number of bad sensors, the trial is then repaired by interpolation or by excluding it from subsequent analysis. All steps of the algorithm are fully automated thus lending itself to the name Autoreject. In order to assess the practical significance of the algorithm, we conducted extensive validation and comparisons with state-of-the-art methods on four public datasets containing MEG and EEG recordings from more than 200 subjects. The comparisons include purely qualitative efforts as well as quantitatively benchmarking against human supervised and semi-automated preprocessing pipelines. The algorithm allowed us to automate the preprocessing of MEG data from the Human Connectome Project (HCP) going up to the computation of the evoked responses. The automated nature of our method minimizes the burden of human inspection, hence supporting scalability and reliability demanded by data analysis in modern neuroscience. (10.1016/j.neuroimage.2017.06.030)
  DOI : 10.1016/j.neuroimage.2017.06.030
• Cournot-Nash Equilibria for Bandwidth Allocation under Base-Station Cooperation
  
  • Gomez J S
  • Vergne A
  • Martins P
  • Decreusefond Laurent
  • Chen Wei
  , 2017. —In this paper, a novel resource allocation scheme based on discrete Cournot-Nash equilibria and optimal transport theory is proposed. The originality of this framework lies in the joint optimization of downlink bandwidth allocation and cooperation between base stations. A tractable formalization is given in the form of a quadratic optimization problem. A low complexity approximate solution is derived and theoretically characterized. Simulations highlight the existence of an optimal working point, that maximizes user satisfaction ratio and network load. The impact of the network deployment on the optimum is numerically investigated, thanks to the β-Ginibre model. Indeed, base stations are assumed to be drawn according to β-Ginibre point processes. Numerical analysis shows that the network performance increases with β going to one.
• Large Scale Density-friendly Graph Decomposition via Convex Programming
  
  • Danisch Maximilien
  • Chan T-H. Hubert
  • Sozio Mauro
  , 2017.
• Robust dynamic range computation for high dynamic range content
  
  • Hulusic Vedad
  • Valenzise Giuseppe
  • Debattista Kurt
  • Dufaux Frederic
  , 2017. High dynamic range (HDR) imaging has become an important topic in both academic and industrial domains. Nevertheless, the concept of dynamic range (DR), which underpins HDR, and the way it is measured are still not clearly understood. The current approach to measure DR results in a poor correlation with perceptual scores (r ≈ 0.6). In this paper, we analyze the limitations of the existing DR measure, and propose several options to predict more accurately subjective DR judgments. Compared to the traditional DR estimates, the proposed measures show significant improvements in Spearman's and Pearson's correlations with subjective data (up to r ≈ 0.9). Despite their straightforward nature, these improvements are particularly evident in specific cases, where the scores obtained by using the classical measure have the highest error compared to the perceptual mean opinion score.
• Very high resolution and interferometric SAR: Markovian and patch-based non-local mathematical models
  
  • Deledalle Charles-Alban
  • Denis Loïc
  • Ferraioli Giampaolo
  • Pascazio Vito
  • Schirinzi Gilda
  • Tupin Florence
  , 2017. This chapter is dedicated to very high resolution (VHR) SAR imagery, including interferometric applications. First, the principles of SAR data acquisition are presented as well as the different types of configurations. The widely adopted Gaussian complex model of fully developed speckle is described as well as more advanced statistical models for VHR SAR data that account for textures. The following two parts are devoted to SAR image estimation and to image denoising within two different frameworks. First, Markovian modeling is introduced and the associated optimization approaches are presented, including graph-cut based optimization. The second framework is the patch-based non-local modeling of SAR complex data. Both frameworks are adapted to SAR images through the use of statistical models specific to SAR imagery. Their applications to amplitude data, interferometry, and fusion with optical data are illustrated. A special focus is given to phase unwrapping applied to single and multi- channel interferometry, showing the usefulness of local and global contextual models. (10.1007/978-3-319-66330-2)
  DOI : 10.1007/978-3-319-66330-2
• Planck intermediate results - L. Evidence of spatial variation of the polarized thermal dust spectral energy distribution and implications for CMB B-mode analysis
  
  • Aghanim N.
  • Ashdown M.
  • Aumont J.
  • Baccigalupi C.
  • Ballardini M.
  • Banday A.J.
  • Barreiro R.B.
  • Bartolo N.
  • Basak S.
  • Benabed K.
  • Bernard Jean-Paul
  • Bersanelli M.
  • Bielewicz P.
  • Bonaldi A.
  • Bonavera L.
  • Bond J.R.
  • Borrill J.
  • Bouchet F.R.
  • Boulanger F.
  • Bracco A.
  • Burigana C.
  • Calabrese E.
  • Cardoso J.F.
  • Chiang H.C.
  • Colombo L.P.L.
  • Combet C.
  • Comis B.
  • Crill B.P.
  • Curto A.
  • Cuttaia F.
  • Davis R.J.
  • de Bernardis P.
  • de Rosa A.
  • de Zotti G.
  • Delabrouille J.
  • Delouis J. M.
  • Di Valentino E.
  • Dickinson C.
  • Diego J.M.
  • Dore O.
  • Douspis M.
  • Ducout A.
  • Dupac X.
  • Dusini S.
  • Efstathiou G.
  • Elsner F.
  • Ensslin T.A.
  • Eriksen H.K.
  • Falgarone E.
  • Fantaye Y.
  • Finelli F.
  • Frailis M.
  • Fraisse A.A.
  • Franceschi E.
  • Frolov A.
  • Galeotta S.
  • Galli S.
  • Ganga K.
  • Genova-Santos R.T.
  • Gerbino M.
  • Ghosh T.
  • Giard M.
  • Gonzalez-Nuevo J.
  • Gorski K.M.
  • Gregorio A.
  • Gruppuso A.
  • Gudmundsson J.E.
  • Hansen F.K.
  • Helou G.
  • Herranz D.
  • Hivon E.
  • Huang Z.
  • Jaffe A.H.
  • Jones W.C.
  • Keihanen E.
  • Keskitalo R.
  • Kisner T.S.
  • Krachmalnicoff N.
  • Kunz M.
  • Kurki-Suonio H.
  • Lagache G.
  • Lahteenmaki A.
  • Lamarre J.M.
  • Lasenby A.
  • Lattanzi M.
  • Lawrence C.R.
  • Le Jeune M.
  • Levrier F.
  • Liguori M.
  • Lilje P.B.
  • Lopez-Caniego M.
  • Lubin P.M.
  • Macias-Perez J.F.
  • Maggio G.
  • Maino D.
  • Mandolesi N.
  • Mangilli A.
  • Maris M.
  • Martin P.G.
  • Martinez-Gonzalez E.
  • Matarrese S.
  • Mauri N.
  • Mcewen J.D.
  • Melchiorri A.
  • Mennella A.
  • Migliaccio M.
  • Mitra S.
  • Miville-Deschenes M.A.
  • Molinari D.
  • Moneti A.
  • Montier L.
  • Morgante G.
  • Moss A.
  • Naselsky P.
  • Norgaard-Nielsen H.U.
  • Oxborrow C.A.
  • Pagano L.
  • Paoletti D.
  • Partridge B.
  • Patrizii L.
  • Perdereau O.
  • Perotto L.
  • Pettorino V.
  • Piacentini F.
  • Plaszczynski S.
  • Polenta G.
  • Puget J.L.
  • Rachen J.P.
  • Reinecke M.
  • Remazeilles M.
  • Renzi A.
  • Rocha G.
  • Rossetti M.
  • Roudier G.
  • Rubino-Martin J.A.
  • Ruiz-Granados B.
  • Salvati L.
  • Sandri M.
  • Savelainen M.
  • Scott D.
  • Sirignano C.
  • Sirri G.
  • Stanco L.
  • Suur-Uski A.S.
  • Tauber J.A.
  • Tenti M.
  • Toffolatti L.
  • Tomasi M.
  • Tristram M.
  • Trombetti T.
  • Valiviita J.
  • Vansyngel F.
  • van Tent F.
  • Vielva P.
  • Wandelt B.D.
  • Wehus I.K.
  • Zacchei A.
  • Zonca A.
  Astronomy & Astrophysics - A&A, EDP Sciences, 2017, 599, pp.A51. The characterization of the Galactic foregrounds has been shown to be the main obstacle in thechallenging quest to detect primordial B-modes in the polarized microwave sky. We make use of the Planck-HFI 2015 data release at high frequencies to place new constraints on the properties of the polarized thermal dust emission at high Galactic latitudes. Here, we specifically study the spatial variability of the dust polarized spectral energy distribution (SED), and its potential impact on the determination of the tensor-to-scalar ratio, r. We use the correlation ratio of the CBBℓ angular power spectra between the 217 and 353 GHz channels as a tracer of these potential variations, computed on different high Galactic latitude regions, ranging from 80% to 20% of the sky. The new insight from Planck data is a departure of the correlation ratio from unity that cannot be attributed to a spurious decorrelation due to the cosmic microwave background, instrumental noise, or instrumental systematics. The effect is marginally detected on each region, but the statistical combination of all the regions gives more than 99% confidence for this variation in polarized dust properties. In addition, we show that the decorrelation increases when there is a decrease in the mean column density of the region of the sky being considered, and we propose a simple power-law empirical model for this dependence, which matches what is seen in the Planck data. We explore the effect that this measured decorrelation has on simulations of the BICEP2-Keck Array/Planck analysis and show that the 2015 constraints from these data still allow a decorrelation between the dust at 150 and 353 GHz that is compatible with our measured value. Finally, using simplified models, we show that either spatial variation of the dust SED or of the dust polarization angle are able to produce decorrelations between 217 and 353 GHz data similar to the values we observe in the data. Key words: cosmic background radiation / cosmology: observations / submillimeter: ISM / dust, extinction⋆ Corresponding author: L. Montier, e-mail: Ludovic.Montier@irap.omp.eu; J. Aumont, e-mail: jonathan.aumont@ias.u-psud.fr (10.1051/0004-6361/201629164)
  DOI : 10.1051/0004-6361/201629164
• Parallel Combining: Making Use of Free Cycles
  
  • Aksenov Vitaly
  • Kuznetsov Petr
  Computing Research Repository, ACM / ArXiv, 2017, abs/1710.07588.
• Classification of MRI data using deep learning and Gaussian process-based model selection
  
  • Bertrand Hadrien
  • Perrot Matthieu
  • Ardon Roberto
  • Bloch Isabelle
  , 2017, pp.745-748. The classification of MRI images according to the anatomical field of view is a necessary task to solve when faced with the increasing quantity of medical images. In parallel, advances in deep learning makes it a suitable tool for computer vision problems. Using a common architecture (such as AlexNet) provides quite good results, but not sufficient for clinical use. Improving the model is not an easy task, due to the large number of hyper-parameters governing both the architecture and the training of the network, and to the limited understanding of their relevance. Since an exhaustive search is not tractable, we propose to optimize the network first by random search, and then by an adaptive search based on Gaussian Processes and Probability of Improvement. Applying this method on a large and varied MRI dataset, we show a substantial improvement between the baseline network and the final one (up to 20% for the most difficult classes).
• Optimal two-step prediction in regression
  
  • Chételat Didier
  • Lederer Johannes
  • Salmon Joseph
  Electronic Journal of Statistics, Shaker Heights, OH : Institute of Mathematical Statistics, 2017, 11 (1), pp.2519-2546.