HAL Portal UTC Université de Technologie de Compiègne
Not a member yet
11652 research outputs found
Sort by
The effect of rhamnolipids on fungal membrane models as described by their interactions with phospholipids and sterols: An in silico study
No full textInternational audienceIntroduction: Rhamnolipids (RLs) are secondary metabolites naturally produced by bacteria of the genera Pseudomonas and Burkholderia with biosurfactant properties. A specific interest raised from their potential as biocontrol agents for crop culture protection in regard to direct antifungal and elicitor activities. As for other amphiphilic compounds, a direct interaction with membrane lipids has been suggested as the key feature for the perception and subsequent activity of RLs.Methods: Molecular Dynamics (MD) simulations are used in this work to provide an atomistic description of their interactions with different membranous lipids and focusing on their antifungal properties.Results and discussion: Our results suggest the insertion of RLs into the modelled bilayers just below the plane drawn by lipid phosphate groups, a placement that is effective in promoting significant membrane fluidification of the hydrophobic core. This localization is promoted by the formation of ionic bonds between the carboxylate group of RLs and the amino group of the phosphatidylethanolamine (PE) or phosphatidylserine (PS) headgroups. Moreover, RL acyl chains adhere to the ergosterol structure, forming a significantly higher number of van der Waals contact with respect to what is observed for phospholipid acyl chains. All these interactions might be essential for the membranotropic-driven biological actions of RLs
Evidence-based data mining method to reveal similarities between materials based on physical mechanisms
No full textInternational audienceMeasuring the similarity between materials is essential for estimating their properties and revealing the associated physical mechanisms. However, current methods for measuring the similarity between materials rely on theoretically derived descriptors and parameters fitted from experimental or computational data, which are often insufficient and biased. Furthermore, outliers and data generated by multiple mechanisms are usually included in the dataset, making the data-driven approach challenging and mathematically complicated. To overcome such issues, we apply the Dempster–Shafer theory to develop an evidential regression-based similarity measurement (eRSM) method, which can rationally transform data into evidence. It then combines such evidence to conclude the similarities between materials, considering their physical properties. To evaluate the eRSM, we used two material datasets, including 3[Formula: see text] transition metal–4[Formula: see text] rare-earth binary and quaternary high-entropy alloys with target properties, Curie temperature, and magnetization. Based on the information obtained on the similarities between the materials, a clustering technique is applied to learn the cluster structures of the materials that facilitate the interpretation of the mechanism. The unsupervised learning experiments demonstrate that the obtained similarities are applicable to detect anomalies and appropriately identify groups of materials whose properties correlate differently with their compositions. Furthermore, significant improvements in the accuracies of the predictions for the Curie temperature and magnetization of the quaternary alloys are obtained by introducing the similarities, with the reduction in mean absolute errors of 36% and 18%, respectively. The results show that the eRSM can adequately measure the similarities and dissimilarities between materials in these datasets with respect to mechanisms of the target properties
Auto-apprentissage à l'aide de prédicteurs de Venn-Abers
No full textNational audienceDans les problèmes d'apprentissage supervisé, il est courant de disposer d'un grand nombre de données non étiquetées, mais de peu de données étiquetées. Il est alors souhaitable d'exploiter les données non étiquetées pour améliorer la procédure d'apprentissage. L'un des moyens pour y parvenir consiste à demander à un modèle de prédire des « pseudo-étiquettes » pour les données non étiquetées, afin de les utiliser pour l'apprentissage. Dans le cadre de l'auto-apprentissage, les pseudo-étiquettes sont fournies par le même modèle que celui qui les exploite. Comme ces pseudo-étiquettes sont par nature incertaines et seulement partiellement fiables, il est naturel de tenir compte de l'incertitude d'étiquetage dans le processus d'apprentissage, ne serait-ce que pour renforcer la procédure d'auto-apprentissage. Cet article décrit une telle approche, dans laquelle nous utilisons des prédicteurs Venn-Abers pour produire des étiquettes crédales calibrées afin de quantifier l'incertitude d'étiquetage. Ces étiquettes sont ensuite intégrées au processus d'apprentissage au moyen d'une fonction de cout adaptée. Les expériences montrent que la prise en compte de l'incertitude d'étiquetage renforce la procédure d'auto-apprentissage et lui permet généralement de converger plus rapidement
Learning Calibrated Belief Functions from Conformal Predictions
No full textInternational audienceWe consider the problem of supervised classification. We focus on the problem of calibrating the classifier's outputs. We show that the p-values provided by Inductive Conformal Prediction (ICP) can be interpreted as a possibility distribution over the set of classes. This allows us to use ICP to compute a predictive belief function which is calibrated by construction. We also propose a learning method which provides p-values in a simpler and faster way, by making use of a multi-output regression model. Results obtained on the Cifar10 and Digits data sets show that our approach is comparable to standard ICP in terms of accuracy and calibration, while offering a reduced complexity and avoiding the use of a calibration set
Quantifying Prediction Uncertainty in Regression Using Random Fuzzy Sets: The ENNreg Model
No full textInternational audienceWe introduce a neural network model for regression in which prediction uncertainty is quantified by Gaussian random fuzzy numbers (GRFNs), a newly introduced family of random fuzzy subsets of the real line that generalizes both Gaussian random variables and Gaussian possibility distributions. The output GRFN is constructed by combining GRFNs induced by prototypes using a combination operator that generalizes Dempster's rule of Evidence Theory. The three output units indicate the most plausible value of the response variable, variability around this value, and epistemic uncertainty. The network is trained by minimizing a loss function that generalizes the negative log-likelihood. Comparative experiments show that this method is competitive, both in terms of prediction accuracy and calibration error, with state-of-the-art techniques such as random forests or deep learning with Monte Carlo dropout. In addition, the model outputs a predictive belief function that can be shown to be calibrated, in the sense that it allows us to compute conservative prediction intervals with specified belief degree
A sequential use of an analytical and a finite element method to compare the magnetic pressures of the SVPWM with DPWM2
No full textInternational audienc
Proposal of a Combined AHP-PROMETHEE Decision Support Tool for Selecting Sustainable Machining Process Based on Toolpath Strategy and Manufacturing Parameters
No full textInternational audienceSustainable manufacturing technologies are the new challenge faced by enterprises, industries, and researchers. The development of a sustainability-based assessment method considering the environmental and economic impacts is crucial to realize viable manufacturing. However, few studies have addressed environmental economics and social flows using a common perspective. Mechanical machining is one of the most-used manufacturing techniques. The overall ecological, economic, and social footprint requires accurate and effective estimation and optimization. Several studies have addressed this issue by examining the entire process of machining, but sustainability flows for machining parameters and toolpaths have remained relatively unexplored. The lack of systematic assistance tools bridging the gap between decision-maker preferences and the three sustainability pillars—economic, social, and environmental—has impeded the widespread adoption of sustainable machining practices. To this end, this paper proposes an integrated approach to the decision-making problem that combines the Analytical Hierarchy Process (AHP) with the Preference Ranking Organization Method for Enrichment Evaluations (PROMETHEE) for selecting a sustainable machining strategy. The sustainability criteria are driven by manufacturing process parameters commonly employed and regulated during the manufacturing phase. This includes toolpath strategies as a qualitative input factor and manufacturing parameters such as cutting speed, feed rate, depth of cut, and stepover as quantitative input factors, affirming the practical applicability of the method in industrial contexts. New fundamental methods are also presented for selecting the most efficient machining parameters and toolpaths according to the weights assigned to each ecological, social, and economic footprint by the decision-maker (the manufacturer or production manager). In this way, sustainable machining strategies in the manufacturing industry will be strengthened in integrity. In a case study of part-end milling, both manufacturing parameters and toolpath strategies are considered to establish sustainable feature-based machining decisions
Collaborative Grid Mapping for Moving Object Tracking Evaluation
No full textInternational audiencePerception of other road users is a crucial task for intelligent vehicles. Perception systems can use on-board sensors only or be in cooperation with other vehicles or with roadside units. In any case, the performance of perception systems has to be evaluated against ground-truth data, which is a particularly tedious task and requires numerous manual operations. In this article, we propose a novel semi-automatic method for pseudo ground-truth estimation. The principle consists in carrying out experiments with several vehicles equipped with LiDAR sensors and with fixed perception systems located at the roadside in order to collaboratively build reference dynamic data. The method is based on grid mapping and in particular on the elaboration of a background map that holds relevant information that remains valid during a whole dataset sequence. Data from all agents is converted in time-stamped observations grids. A data fusion method that manages uncertainties combines the background map with observations to produce dynamic reference information at each instant. Several datasets have been acquired with three experimental vehicles and a roadside unit. An evaluation of this method is finally provided in comparison to a handmade ground truth
Evaluation of the impact of urban wet weather discharges (UWWD) on the Seine River by high resolution mass spectrometry (HRMS) and development of predictive models
No full textInternational audienceThe impact of urban wet weather discharges (UWWD) from combined sewer overflows on water quality is a major environmental concern and has been associated with an increase in ecotoxic responses from bioassays performed as part of a monitoring program of the Seine River. UWWD contain a diversity of micropollutants that can originate from both runoff and wastewater overflow, and thus are difficult to characterize. In this study, a non-targeted approach using high-resolution mass spectrometry (HRMS) was employed to characterize the presence of UWWD contaminants in surface water. Various amounts of UWWD were added to river water samples and analyses were performed by ultraperformance liquid chromatography coupled to ion mobility spectrometry and QTOF detection, after solid-phase extraction. Statistical analyses were performed on the HRMS signals to first determine the smallest amount of UWWD that would cause a significant difference to the HRMS signals of the surface water. Both supervised and unsupervised analysis techniques were then used to develop two types of predictive models: regression models designed to predict the proportion of UWWD in a water sample, and classification models designed to distinguish whether the sample is lightly or heavily contaminated with UWWD. This methodology is an important step towards environmental monitoring and water quality management, especially during periods of intense rainfall. The developed predictive models could be used to efficiently assess the presence and amount of UWWD in surface water, without targeting specific contaminants. Furthermore, the results of this study pave the way for an integration of these approaches with ecotoxicological analyses for a more comprehensive assessment of the environmental impact of UWWD
Hidden markov models in reliability and maintenance
No full textInternational audienceAlthough the hidden Markov models (HMM) are very popular in many applied areas their use in reliability engineering is limited. Problems such as the selection of the HMM model by choosing the appropriatenumber of states, or problems of prediction of failures have not been widely covered in the literature.This paper is concerned with the use of HMMs where the state of the system is not directly observableand instead certain indicators of the true situation are provided via a control system. A hidden modelcan provide key information about the system dependability such as the failed component of the system, the reliability of the system and related measures. A maximum-likelihood estimator of the systemreliability is obtained and its asymptotic properties are studied. Finally, the maintenance of the systemis considered in this context and new preventive maintenance strategies are defined and their efficiencyis measured in terms of expected cost. To prove the finite sample performance of the methodology, anextensive simulation study is developed