Université de Bretagne Sud

Domain adaptation is one of the most challenging tasks of modern data analytics. If the adaptation is done correctly, models built on a specific data representation become more robust when confronted to data depicting the same classes, but described by another observation system. Among the many strategies proposed, finding domain-invariant representations has shown excellent properties, in particular since it allows to train a unique classifier effective in all domains. In this paper, we propose a regularized unsupervised optimal transportation model to perform the alignment of the representations in the source and target domains. We learn a transportation plan matching both PDFs, which constrains labeled samples of the same class in the source domain to remain close during transport. This way, we exploit at the same time the labeled samples in the source and the distributions observed in both domains. Experiments on toy and challenging real visual adaptation examples show the interest of the method, that consistently outperforms state of the art approaches. In addition, numerical experiments show that our approach leads to better performances on domain invariant deep learning features and can be easily adapted to the semi-supervised case where few labeled samples are available in the target domain.

Additive manufacturing and digital fabrication bring new horizons to concrete and cement-based material construction. 3D printing inspired construction techniques that have recently been developed at laboratory scale for cement-based materials. This study aims to investigate the role of the structural build-up properties of cement-based materials in such a layer by layer construction technique. As construction progresses, the cement-based materials become harder with time. The mechanical strength of the cement-based materials must be sufficient to sustain the weight of the layers subsequently deposited. It follows that the comparison of the mechanical strength, which evolves with time (i.e. structural build-up), with the loading due to layers subsequently deposited, can be expected to provide the optimal rate of layer by layer construction. A theoretical framework has been developed to propose a method of optimization of the building rate, which is experimentally validated in a layer-wise built column.

This article offers a comprehensive overview of the underlying physics relevant to an understanding of materials processing during the various production steps in extrusion-based 3D concrete printing (3DCP). Understanding the physics governing the processes is an important step towards the purposeful design and optimization of 3DCP systems as well as their efficient and robust process control. For some processes, analytical formulas based on the relevant physics have already enabled reasonable predictions with respect to material flow behavior and buildability, especially in the case of relatively simple geometries. The existing research in the field was systematically compiled by the authors in the framework of the activities of the RILEM Technical Committee 276 “Digital fabrication with cement-based materials”. However, further research is needed to develop reliable tools for the quantitative analysis of the entire process chain. To achieve this, experimental efforts for the characterization of material properties need to go hand in hand with comprehensive numerical simulation.

High-level synthesis raises the design abstraction level and allows rapid generation of optimized RTL hardware for performance, area, and power requirements. This article gives an overview of state-of-the-art HLS techniques and tools.

International audience

Pseudomonas aeruginosa is a Gram-negative bacterium belonging to the γ-proteobacteria. Like other members of the Pseudomonas genus, it is known for its metabolic versatility and its ability to colonize a wide range of ecological niches, such as rhizosphere, water environments and animal hosts, including humans where it can cause severe infections. Another particularity of P. aeruginosa is its high intrinsic resistance to antiseptics and antibiotics, which is partly due to its low outer membrane permeability. In contrast to Enterobacteria, pseudomonads do not possess general diffusion porins in their outer membrane, but rather express specific channel proteins for the uptake of different nutrients. The major outer membrane 'porin', OprF, has been extensively investigated, and displays structural, adhesion and signaling functions while its role in the diffusion of nutrients is still under discussion. Other porins include OprB and OprB2 for the diffusion of glucose, the two small outer membrane proteins OprG and OprH, and the two porins involved in phosphate/pyrophosphate uptake, OprP and OprO. The remaining nineteen porins belong to the so-called OprD (Occ) family, which is further split into two subfamilies termed OccD (8 members) and OccK (11 members). In the past years, a large amount of information concerning the structure, function and regulation of these porins has been published, justifying why an updated review is timely.

In computer vision, one is often confronted with problems of domain shifts,\nwhich occur when one applies a classifier trained on a source dataset to target\ndata sharing similar characteristics (e.g. same classes), but also different\nlatent data structures (e.g. different acquisition conditions). In such a\nsituation, the model will perform poorly on the new data, since the classifier\nis specialized to recognize visual cues specific to the source domain. In this\nwork we explore a solution, named DeepJDOT, to tackle this problem: through a\nmeasure of discrepancy on joint deep representations/labels based on optimal\ntransport, we not only learn new data representations aligned between the\nsource and target domain, but also simultaneously preserve the discriminative\ninformation used by the classifier. We applied DeepJDOT to a series of visual\nrecognition tasks, where it compares favorably against state-of-the-art deep\ndomain adaptation methods.\n

Plastics are ubiquitous in the oceans and constitute suitable matrices for bacterial attachment and growth. Understanding biofouling mechanisms is a key issue to assessing the ecological impacts and fate of plastics in marine environment. In this study, we investigated the different steps of plastic colonization of polyolefin-based plastics, on the first one hand, including conventional low-density polyethylene (PE), additivated PE with pro-oxidant (OXO), and artificially aged OXO (AA-OXO); and of a polyester, poly(3-hydroxybutyrate-co-3-hydroxyvalerate) (PHBV), on the other hand. We combined measurements of physical surface properties of polymers (hydrophobicity and roughness) with microbiological characterization of the biofilm (cell counts, taxonomic composition, and heterotrophic activity) using a wide range of techniques, with some of them used for the first time on plastics. Our experimental setup using aquariums with natural circulating seawater during 6 weeks allowed us to characterize the successive phases of primo-colonization, growing, and maturation of the biofilms. We highlighted different trends between polymer types with distinct surface properties and composition, the biodegradable AA-OXO and PHBV presenting higher colonization by active and specific bacteria compared to non-biodegradable polymers (PE and OXO). Succession of bacterial population occurred during the three colonization phases, with hydrocarbonoclastic bacteria being highly abundant on all plastic types. This study brings original data that provide new insights on the colonization of non-biodegradable and biodegradable polymers by marine microorganisms.

In a way similar to the string-to-string correction problem, we address discrete time series similarity in light of a time-series-to-time-series-correction problem for which the similarity between two time series is measured as the minimum cost sequence of edit operations needed to transform one time series into another. To define the edit operations, we use the paradigm of a graphical editing process and end up with a dynamic programming algorithm that we call Time Warp Edit Distance (TWED). TWED is slightly different in form from Dynamic Time Warping (DTW), Longest Common Subsequence (LCSS), or Edit Distance with Real Penalty (ERP) algorithms. In particular, it highlights a parameter that controls a kind of stiffness of the elastic measure along the time axis. We show that the similarity provided by TWED is a potentially useful metric in time series retrieval applications since it could benefit from the triangular inequality property to speed up the retrieval process while tuning the parameters of the elastic measure. In that context, a lower bound is derived to link the matching of time series into downsampled representation spaces to the matching into the original space. The empiric quality of the TWED distance is evaluated on a simple classification task. Compared to Edit Distance, DTW, LCSS, and ERP, TWED has proved to be quite effective on the considered experimental task.

In this paper, we address the issue of recovering and segmenting the apparent velocity field in sequences of images. As for motion estimation, we minimize an objective function involving two robust terms. The first one cautiously captures the optical flow constraint, while the second (a priori) term incorporates a discontinuity-preserving smoothness constraint. To cope with the nonconvex minimization problem thus defined, we design an efficient deterministic multigrid procedure. It converges fast toward estimates of good quality, while revealing the large discontinuity structures of flow fields. We then propose an extension of the model by attaching to it a flexible object-based segmentation device based on deformable closed curves (different families of curve equipped with different kinds of prior can be easily supported). Experimental results on synthetic and natural sequences are presented, including an analysis of sensitivity to parameter tuning.

Recently, interest has been increasing in natural fibres as composite reinforcing fillers for polymer-based filaments manufactured with the Fused Deposition Modeling (FDM) process, despite their moderate mechanical properties compared to pure polymer. An innovative way was proposed in the present work to optimize the mechanical properties of biocomposites. It was based on novel continuous flax fibre/PLA (cFF/PLA) composite filaments made with a customized co-extrusion process and printed with a simple and affordable printing machine. The microstructure of the printed cFF/PLA biocomposite evidenced a homogeneous distribution of yarn within the cross section, while the twisted flax yarn led to fibre-rich areas at mesoscale. The cFF/PLA showed tensile modulus and strength values that exceeded the only available published result on continuous natural fibre printed composites by >4.5 times. Tensile properties were in the same range as those for continuous glass fibre/PolyAmide (PA) printed composites, paving the way for the use of biocomposites in structural applications. Their weakest point was their transverse properties that remained poorer than similar flax/PLA thermocompressed composites.

Background: We conducted a meta-analysis in nonmetastatic breast cancer patients treated by neoadjuvant chemotherapy (NCT) to assess the clinical validity of circulating tumor cell (CTC) detection as a prognostic marker. Methods: We collected individual patient data from 21 studies in which CTC detection by CellSearch was performed in early breast cancer patients treated with NCT. The primary end point was overall survival, analyzed according to CTC detection, using Cox regression models stratified by study. Secondary end points included distant disease-free survival, locoregional relapse-free interval, and pathological complete response. All statistical tests were two-sided. Results: Data from patients were collected before NCT (n = 1574) and before surgery (n = 1200). CTC detection revealed one or more CTCs in 25.2% of patients before NCT; this was associated with tumor size (P < .001). The number of CTCs detected had a detrimental and decremental impact on overall survival (P < .001), distant disease-free survival (P < .001), and locoregional relapse-free interval (P < .001), but not on pathological complete response. Patients with one, two, three to four, and five or more CTCs before NCT displayed hazard ratios of death of 1.09 (95% confidence interval [CI] = 0.65 to 1.69), 2.63 (95% CI = 1.42 to 4.54), 3.83 (95% CI = 2.08 to 6.66), and 6.25 (95% CI = 4.34 to 9.09), respectively. In 861 patients with full data available, adding CTC detection before NCT increased the prognostic ability of multivariable prognostic models for overall survival (P < .001), distant disease-free survival (P < .001), and locoregional relapse-free interval (P = .008). Conclusions: CTC count is an independent and quantitative prognostic factor in early breast cancer patients treated by NCT. It complements current prognostic models based on tumor characteristics and response to therapy.

The composition, size distribution, and abundance of floating plastic debris in surface waters of the Mediterranean Sea were analyzed in relation to distance to land. We combined data from previously published reports with an intensive sampling in inshore waters of the Northwestern Mediterranean. The highest plastic concentrations were found in regions distant from from land as well as in the first kilometer adjacent to the coastline. In this nearshore water strip, plastic concentrations were significantly correlated with the nearness to a coastal human population, with local areas close to large human settlements showing hundreds of thousands of plastic pieces per km2. The ratio of plastic to plankton abundance reached particularly high values for the coastal surface waters. Polyethylene, polypropylene and polyamides were the predominant plastic polymers at all distances from coast (86 to 97% of total items), although the diversity of polymers was higher in the 1-km coastal water strip due to a higher frequency of polystyrene or polyacrylic fibers. The plastic size distributions showed a gradual increase in abundance toward small sizes indicating an efficient removal of small plastics from the surface. Nevertheless, the relative abundance of small fragments (< 2 mm) was higher within the 1-km coastal water strip, suggesting a rapid fragmentation down along the shoreline, likely related with the washing ashore on the beaches. This study constitutes a first attempt to determine the impact of plastic debris in areas closest to Mediterranean coast. The presence of a high concentration of plastic including tiny plastic items could have significant environmental, health and economic impacts.

To date, the literature has focused on synthetic fibre-reinforced composites, but it has not adequately addressed the unique properties that differentiate natural from synthetic fibres, such as their natural variation in microstructure and composition across species. This review paper proposes a critical overview of the current state of 3D printing of natural fibre-reinforced composites or biocomposites for mechanical purposes, as well as an overview of their role in 4D printing for stimuli-responsive applications. The paper is structured as follows: after the first part recalling the specificities of natural fibres and their associated composites, the two main sections are each divided into two parts presenting an analysis of the available data to provide fundamental understandings and a discussion and outlook for the future. Natural discontinuous fibre-reinforced polymers exhibit moderate mechanical properties compared to composites manufactured by conventional processes due to specific factors of the 3D printing process, such as high porosity, low fibre content, and a very low fibre-aspect ratio (L/d). Hygromorph BioComposites (HBC) are categorized into a new class of smart materials that could be used for 4D printing of shape-changing mechanisms. Fibre content, fibre orientation control, and fibre continuity are outlined in relation to known challenges in actuation performance.

We are presenting a new method of processing polystyrene−silica nanocomposites, which results in a very well-defined dispersion of small primary aggregates (assembly of 15 nanoparticles of 10 nm diameter) in the matrix. The process is based on the use of a high boiling point solvent, in which the nanoparticles are well dispersed, and a controlled evaporation procedure. The filler’s fine network structure is determined over a wide range of sizes, using a combination of small-angle neutron scattering (SANS) and transmission electronic microscopy (TEM) experiments. The mechanical response of the nanocomposite material has been investigated for both small (ARES oscillatory shear and dynamical mechanical analysis) and large deformations (uniaxial traction) as a function of the concentration of the particles in the matrix. Our findings show that with a simple tuning parameter, the silica filler volume fraction, we can investigate in the same way the structure−property correlations related to the two main reinforcement effects: the filler network contribution and a filler−polymer matrix effect. Above a silica volume fraction threshold, we were able to highlight a divergence of the reinforcement factor, which is clearly correlated to the formation of a connected network built up from the finite-size primary aggregates and is thus a direct illustration of the filler network contribution. For a silica volume fraction lower than this percolation threshold, we obtain a new additional elastic contribution of the material, of longer terminal time than the matrix. This cannot be attributed to the filler network effect, as the filler is well dispersed, each element separated from the next by a mean distance of 60 nm. This new result, which implies the filler−matrix contribution of the reinforcement, must include interfacial contributions. Nevertheless, it cannot be described solely with the concept of glassy layer, i.e., only as a dynamic effect, because its typical length scale extension should be much shorter, of the order of 2 nm. This implies a need to reconsider the polymer−filler interaction potential and to take into account a possible additional polymer conformational contribution due to the existence of indirect long-range bridging of the filler by the chains.

ABSTRACT The publication of sustainability reports has increased significantly in most western countries in recent years. The fact that this type of reporting is on a voluntary basis in a number of countries raises questions about whether capital markets take these reports into account. This study attempts to address this question, drawing on a sample of Canadian companies listed on the Toronto Stock Exchange. The results suggest that investors positively value this type of reporting. These findings support the relevance of initiatives like the Global Reporting Initiative, the UN Global Compact, and that launched by the International Organization for Standardization (ISO), which focus on the development of recognized guidelines for sustainability reporting. Our findings can also serve as arguments to facilitate a firm's voluntary commitment to such types of disclosures. Copyright © 2012 John Wiley &amp; Sons, Ltd and ERP Environment.

Multispectral image pairs can provide complementary visual information, making pedestrian detection systems more robust and reliable. To benefit from both RGB and thermal IR modalities, we introduce a novel attentive multispectral feature fusion approach. Under the guidance of the inter- and intra-modality attention modules, our deep learning architecture learns to dynamically weigh and fuse the multispectral features. Experiments on two public multi-spectral object detection datasets demonstrate that the proposed approach significantly improves the detection accuracy at a low computation cost.

The work presented here aims to study and compare the performance of a polyvinylidene fluoride (PVDF) electrospun membrane, unmodified cellulose nanofiber (CNF) based PVDF membrane, and Meldrum’s acid (2,2-dimethyl-1,3-dioxane-4,6-dione) modified CNF-based PVDF membranes against the Fe2O3 nanoparticle filtration and crystal violet (CV) dye adsorption. Herein, we introduced a facile method to produce a unique green adsorbent material from cellulose nanofibers (CNFs) via a nonsolvent assisted procedure using Meldrum’s acid as an esterification agent to enhance the adsorption toward positively charged crystal violet dyes. Most of the surface modifications of cellulose nanofibers have been done using toxic organic solvents like pyridine, dimethyl acetate, (2,2,6,6-tetramethylpiperidin-1-yl)oxyl (TEMPO), etc. So far, this is the first report on the surface modification of cellulose nanofibers via a nonsolvent assisted procedure. Both CNF-based PVDF membranes were prepared by successive coating of modified and unmodified CNFs on to the surface of a PVDF electrospun membrane. All the demonstrated membranes showed high filtration capacity against the Fe2O3 nanoparticles. With the 10 mg/L of crystal violet (CV) aqueous solution, CV adsorption of PVDF electrospun membrane, and unmodified CNF-based PVDF membrane was around 1.368 and 2.948 mg/g of the membrane respectively, whereas it was 3.984 mg/g of the membrane by Meldrum’s acid CNF-based PVDF membrane. The demonstrated Meldrum’s acid modified CNF-based PVDF membrane was proven to be the efficient media that can concurrently eliminate the Fe2O3 nanoparticles and CV dyes from the water. The investigation into the surface chemistries of cellulose nanofibers beyond the adoption of toxic solvents can enhance the economic usefulness of the process and also yield a new ecofriendly adsorbent material that is agreeable to adsorbing various toxic pollutants.

Dermatological and cosmetics fields have recently started to focus on the human skin microbiome and microbiota, since the skin microbiota is involved in the health and dysbiosis of the skin ecosystem. Amongst the skin microorganisms, Staphylococcus epidermidis and Cutibacterium acnes, both commensal bacteria, appear as skin microbiota sentinels. These sentinels have a key role in the skin ecosystem since they protect and prevent microbiota disequilibrium by fighting pathogens and participate in skin homeostasis through the production of beneficial bacterial metabolites. These bacteria adapt to changing skin microenvironments and can shift to being opportunistic pathogens, forming biofilms, and thus are involved in common skin dysbiosis, such as acne or atopic dermatitis. The current evaluation methods for cosmetic active ingredient development are discussed targeting these two sentinels with their assets and limits. After identification of these objectives, research of the active cosmetic ingredients and products that maintain and promote these commensal metabolisms, or reduce their pathogenic forms, are now the new challenges of the skincare industry in correlation with the constant development of adapted evaluation methods.

Education for Sustainable Development (ESD) is a concept that evolves in line with emerging sustainability issues. In the 2030 Agenda for Sustainable Development, ESD is embraced in Goal 4, Target 4.7, and reflected in other Goals and Targets. The approach towards the 2030 Agenda is important, not only because of the crucial role that education will play in the implementation of the Goals, but also in increasing its impact by orienting towards the emerging sustainability challenges. Therefore, there is a high demand for research to better understand ESD interactions with the 2030 Agenda framework in specific contexts. This study addresses the implications of Target 4.7 on the 17 Sustainable Development Goals, in the context of the Regional Centres of Expertise on Education for Sustainable Development as global multi-stakeholder networks. It analyses the interaction of Target 4.7 with other Goals and Targets, in order to identify the strongest connections amongst thematic sustainability issues. The findings revealed through statistical analyses and a comprehensive literature review, that the prevailing components of ESD are strongly interconnected with Goals 2, 3, 4, 7, 11, 12, 13, 15. Thus reinforcing that the multidimensional aspects of ESD in relation to the SDGs are stronger regarding the current complex issues such as, education, climate, energy, sustainable cities, natural habitat, consumption and production. Although the nature of multi-stakeholder networks allows for diverse approaches of ESD towards the 2030 Agenda, the study indicates the importance of partnership and informal learning for reflection of global sustainability issues in regional platforms.