Publications

2023

• Rebalancing gradient to improve self-supervised co-training of depth, odometry and optical flow predictions
  
  • Hariat Marwane
  • Manzanera Antoine
  • Filliat David
  , 2023, pp.1267-1276. We present CoopNet, an approach that improves the cooperation of co-trained networks by dynamically adapting the apportionment of gradient, to ensure equitable learning progress. It is applied to motion-aware self-supervised prediction of depth maps, by introducing a new hybrid loss, based on a distribution model of photo-metric reconstruction errors made by, on the one hand the depth + odometry paired networks, and on the other hand the optical flow network. This model essentially assumes that the pixels from moving objects (that must be discarded for training depth and odometry), correspond to those where the two reconstructions strongly disagree. We justify this model by theoretical considerations and experimental evidences. A comparative evaluation on KITTI and CityScapes datasets shows that CoopNet improves or is comparable to the state-of-the-art in depth, odometry and optical flow predictions. Our code is available here: https://github.com/mhariat/CoopNet.
• Computational protein design repurposed to explore enzyme vitality and help predict antibiotic resistance
  
  • Michael Eleni
  • Saint-Jalme Rémy
  • Mignon David
  • Simonson Thomas
  Frontiers in Molecular Biosciences, Frontiers Media, 2023, 9, pp.905588. In response to antibiotics that inhibit a bacterial enzyme, resistance mutations inevitably arise. Predicting them ahead of time would aid target selection and drug design. The simplest resistance mechanism would be to reduce antibiotic binding without sacrificing too much substrate binding. The property that reflects this is the enzyme "vitality", defined here as the difference between the inhibitor and substrate binding free energies. To predict such mutations, we borrow methodology from computational protein design. We use a Monte Carlo exploration of mutation space and vitality changes, allowing us to rank thousands of mutations and identify ones that might provide resistance through the simple mechanism considered. As an illustration, we chose dihydrofolate reductase, an essential enzyme targeted by several antibiotics. We simulated its complexes with the inhibitor trimethoprim and the substrate dihydrofolate. 20 active site positions were mutated, or "redesigned" individually, then in pairs or quartets. We computed the resulting binding free energy and vitality changes. Out of seven known resistance mutations involving active site positions, five were correctly recovered. Ten positions exhibited mutations with significant predicted vitality gains. Direct couplings between designed positions were predicted to be small, which reduces the combinatorial complexity of the mutation space to be explored. It also suggests that over the course of evolution, resistance mutations involving several positions do not need the underlying point mutations to arise all at once: they can appear and become fixed one after the other. (10.3389/fmolb.2022.905588)
  DOI : 10.3389/fmolb.2022.905588
• Key role of boundary conditions for the 2D modeling of crack propagation in linear elastic Compact Tension tests
  
  • Triclot J.
  • Corre T.
  • Gravouil A.
  • Lazarus V.
  Engineering Fracture Mechanics, Elsevier, 2023, 277, pp.109012. In fracture mechanics, the use of experimental tests are fundamental to characterize the material properties in terms of crack initiation and propagation behavior. When modeled in boundary value problems, simplifications need to be made. Notably, the loading has to be reduced to a set of boundary conditions and the choice between plane stress and plane strain has to be done in the 2D case. Here we focus on the Compact Tension (CT) test which is a fracture setup commonly used to measure the fracture toughness at crack propagation onset and we question the possibility to use it to study crack propagation. For this, the tests are monitored by digital image correlation and compared to finite element method simulations. Three ways to guide the choice between plane stress and plane strain hypotheses are proposed. They lead to the same conclusion that the plane stress conditions are the most relevant for the geometry of the samples used here. The key role of boundary conditions is highlighted by testing several models, with imposed force or displacement boundary conditions, against the experimental data. Imposed force boundary conditions on the pin are shown to be able to reproduce the experiments before crack propagation and to be insensitive to the way this force is applied, in line with Saint Venant principle. The results with imposed displacement are in contrary very sensitive to their distribution along the pin. While the stage before propagation is accurately predicted by imposed forces, we show that for the propagation phase, Saint Venant is put in default and accurate results can only be obtained by imposing the displacement fields issued from the digital image correlation. These results can be extended to other fracture experiments, involving pin loading, like the Compact Tension Shear (CTS) or the (Tappered) Double Cantilever Beam ((T)DCB) tests. (10.1016/j.engfracmech.2022.109012)
  DOI : 10.1016/j.engfracmech.2022.109012
• The linear sampling method for random sources
  
  • Garnier Josselin
  • Haddar Houssem
  • Montanelli Hadrien
  SIAM Journal on Imaging Sciences, Society for Industrial and Applied Mathematics, 2023, 16 (3), pp.1572-1593. (10.1137/22M1531336)
  DOI : 10.1137/22M1531336
• Compétitions et performance dans différentes sphères sociales À propos de The Performance Complex de David Stark
  
  • Colin Laure
  Le Libellio d'AEGIS, Libellio d'AEGIS, 2023, 19 (3), pp.5-12. <div><p>la performance a engendré un monde où de plus en plus d'activités de la vie quotidienne sont évaluées et vécues en fonction de leur rendement. Dans cette société, certains produisent et se produisent tandis que d'autres observent et évaluent. De plus en plus d'activités et de domaines de la vie sont ainsi vécus en termes de mesures de la performance (Muller, 2018). Les entreprises surveillent attentivement les performances de leurs employés, les marchés boursiers enregistrent les performances des entreprises, les étudiants évaluent leurs professeurs à la fin de chaque semestre et les clients des restaurants notent leurs repas quotidiennement. L'émergence de cette société où chacun se produit et laisse des traces génère de l'anxiété liée à la position de chacun dans les classements de diverses sphères.</p><p>Selon David Stark, qui a coordonné l'ouvrage collectif intitulé The Performance Complex: Competition and Competitions in Social Life, il convient de désigner cet ensemble de mesures, de réseaux et d'émotions comme étant le complexe de la performance (the performance complex). Pour mieux comprendre l'avènement de ce phénomène, l'ouvrage paru en 2020 rassemble les contributions de nombreux chercheurs oeuvrant dans le domaine émergent des études sociales sur la valorisation (valuation). Les auteurs, tels que</p></div>
• Comment analyser les trajectoires des innovations ? Retour sur les travaux de Bruno Latour
  
  • Colin Laure
  Le Libellio d'AEGIS, Libellio d'AEGIS, 2023, 19 (1), pp.43-51. <div><p>L es études menées par Bruno Latour constituent une base de connaissances permettant aux chercheurs en gestion de dépasser une vision ossifiée des organisations (Czarniawska, 2014 ; 2017). En ignorant les hiérarchies, micro et macro, ses travaux décrivent un monde plat, où les connexions entre des entités hybrides (composées d'humains et non-humains) sont constamment redéfinies. De plus, en combinant la philosophie, la sociologie, l'anthropologie, l'histoire, la sémiologie et même la fiction dans son exploration de modes de travail, sous le label d'anthropologie des sciences et des techniques, Bruno Latour a réussi à imposer une nouvelle manière de comprendre la trajectoire des innovations.</p></div>
• Effect of the deposition direction on fracture propagation in a Duplex Stainless Steel manufactured by Directed Energy Deposition
  
  • Roucou David
  • Corre Thomas
  • Rolland Gilles
  • Lazarus Véronique
  Materials Science and Engineering: A, Elsevier, 2023. Dense volumes of duplex stainless steel are manufactured by directed energy deposition. Compact tension specimens are machined from these volumes in order to evaluate the fracture toughness in two directions : parallel or perpendicular to the deposited layers. Different values are measured in the two cases. In order to understand this anisotropy, additional analyzes are performed on the cracked specimens post-mortem. A classical metallography analysis reveals the highly oriented structure of the material, as well as phase localization. The study of the fracture surface reveals several points. At the macroscale, while the crack surfaces are flat in the parallel case, pronounced shear lips cover half of the fracture surface in the perpendicular case. At the microscale, fracture is ruled by microvoid coalescence. The mesoscale, which is inherited from the deposition strategy, is found to pilot the crack growth. The border between the primary solidified melt pools and the heat-affected zones, which corresponds to the interface between the deposited layers, is the preferred area for crack growth. Analyzing the crack surface roughness confirms the dominance of the mesoscale, as its characteristic lengthscale is retrieved. This explains the differences observed for the two tested directions of fracture: in the parallel case, the crack is aligned with the weak interfaces between layers, which channel the crack growth; in the orthogonal one, out-of-plane excursion of the crack becomes possible allowing the crack to follow a tortuous three-dimensional path that results in a higher toughness than in the parallel situation. (10.1016/j.msea.2023.145176)
  DOI : 10.1016/j.msea.2023.145176
• Simple Postsynthesis Thermal Treatment toward High Luminescence Performance of Rare Earth Vanadate Nanoparticles
  
  • Perrella Rafael Vieira
  • Mohammedi Rabei
  • Kuhner Robin
  • Cardone Christophe
  • Larquet Eric
  • Alexandrou Antigoni
  • de Sousa Filho Paulo Cesar
  • Gacoin Thierry
  Crystal Growth & Design, American Chemical Society, 2023, 23 (8), pp.5389-5396. Optical applications of colloidal oxide nanoparticles are often limited by low luminescence efficiencies caused by poor crystallinity and surface quenching. Bulk oxides prepared via conventional high-temperature annealing offer intense luminescence but commonly fail to yield stable colloidal dispersions. Coupling the best of these two situations to afford highly crystalline, dispersible nanoparticles with luminescence performance exceeding bulk solids is still challenging, thus requiring new safe, scalable, and reproducible methodologies. Herein we report a silicate-coating strategy followed by aggregate elimination to recover stable colloids of 40-150 nm single crystalline rare earth vanadates after unprotected annealing (800-1000 °C). Eu3+-doped nanoparticles showed enhanced photostability and ~50% emission quantum yields in water (λexc=280 nm), while Dy3+-, Tm3+-, and Yb3+/Er3+-doped vanadates provided remarkably intense multicolour emissions via downshift or upconversion luminescence. We correlated spectroscopic properties of pristine and annealed solids to microstructural characteristics to explain the superior emission features, opening new perspectives for sensing applications. (10.1021/acs.cgd.3c00308)
  DOI : 10.1021/acs.cgd.3c00308