Scientific Publications of the University of Toulouse II Le Mirail
Not a member yet
92205 research outputs found
Sort by
A modular framework for the backward error analysis of GMRES
No full textInternational audienceThe Generalized Minimal Residual methods (GMRES) for the solution of general square linear systems is a class of Krylov-based iterative solvers for which there exist backward error analyses that guarantee the computed solution in inexact arithmetic to reach certain attainable accuracies. Unfortunately, these existing backward error analyses cover a relatively small subset of the possible GMRES variants and cannot be used straightforwardly in general to derive new backward error analyses for variants that do not yet have one. We propose a backward error analysis framework for GMRES that simplifies the process of determining error bounds of many existing and future variants of GMRES. This framework describes modular bounds for the attainable normwise backward and forward errors of the computed solution that can be specialized for a given GMRES variant under minimal assumptions. To assess the relevance of our framework we first show that it is compatible with the previous rounding error analyses of GMRES in the sense that it delivers (almost) the same error bounds under (almost) the same conditions. Second, we explain how to use this framework to determine new error bounds for GMRES algorithms that do not have yet or have an incomplete backward error analysis, such as simpler GMRES, CGS2-GMRES, and mixed precision GMRES
"L’enfant et le philosophe : visibilité et invisibilité de l’enfant dans la tradition philosophique."
No full textInternational audienc
Sorry, I Didn’t Get It ! It’s the Emotion : Emotional Influences on L2 Reading Comprehension through Eye-Tracking and Physiological Measures
No full textInternational audienceThis doctoral research investigates how emotions influence the learning of reading comprehension in a foreign language, viewed both as an academic task and a complex cognitive skill. Although the interdependence of emotion and cognition is well established (Altmann et al., 2012; Pessoa, 2008) and studies have highlighted links between affect and text comprehension in L2 (Jian et al., 2025; Roch-Veiras, 2009, 2013, 2015, Ziqing et al., 2025), the nature and mechanisms of these relationships remain unclear. The present study aims to explore how positive and negative emotions affect L2 text comprehension and to what extent these effects can be observed through eye, pupil, and heart activity during reading.The theoretical framework is based on the Resource Allocation Model (RAM) (Ellis & Ashbrook, 1988 ; Ellis & Moore, 1999), which posits that emotional states consume attentional and cognitive resources, potentially disrupting task focus through non-task-related thoughts such as rumination or stress. Within this model, emotions are expected to modulate attentional control and processing efficiency during reading. The central research question is: Do positive (e.g., joy) and negative (e.g., sadness) emotions differentially influence the learning of L2 reading comprehension, and how can these influences be measured in eye-movement, pupillary, and cardiac behaviour during comprehension tasks?The experimental protocol involves 40 adult learners of French as a Foreign Language (B2 CEFR). Participants are exposed to four conditions combining music and reading: three emotional conditions (positive, negative, and neutral music excerpts) and one silent control condition. After listening to the assigned music, each participant reads a neutral narrative text generated with an AI-based text creation tool, which ensures controlled linguistic level, length, and semantic depth across conditions. Comprehension is assessed through multiple-choice questions and two written feedback tasks (immediate and delayed, one week later). Emotional responses are measured via a Likert-scale questionnaire (before and after reading) and through pupillary and cardiac measures. Eye-tracking data provide indicators of reading behaviour and processing effort.The material has been validated through pre-tests and a pilot study. Data collection is currently in progress, and preliminary results are expected to be presented at the conference. Both qualitative and quantitative analyses will be conducted, comparing reading patterns, emotional ratings, and comprehension outcomes across conditions.This study seeks to contribute to a better understanding of the interplay between affect and cognitive processing in L2 reading, combining theoretical and experimental perspective
Dommages par déplacement atomiques induits par les rayons gamma dans des microvolumes de silicium : taux de génération des défauts et signal télégraphique aléatoire.
No full textInternational audienceSingle displacement damage-induced defects created by 60Co gamma-rays are studied in silicon microvolumes using a state-of-the-art CMOS Image Sensor (CIS). The Dark Current Spectroscopy (DCS) technique used in this work enables spatial and temporal tracking of the identified defect generation rates with a single defect sensitivity. The generation rate, energy level, annealing temperature and quantity of expected point defects (including the divacancy V2 and the E-center As-V or P-V) are estimated. An isochronal annealing experiment up to 300°C highlights the creation of new defects at high temperature such as V2OH, VOH and VmOn complexes. This study also demonstrates that 60Co gamma rays can generate bulk defect-induced Dark Current Random Telegraph Signal (DC-RTS) with a very specific temporal and amplitude signature. A divacancy-impurity complex is suspected to be at the origin of this DC-RTS behavior
Preparing relay peasants to become intermediaries in the dissemination of local agroecological knowledge: the case of three West African territories
No full textInternational audienceLong denigrated, peasant knowledge is now seen as a stock of knowledge potentially available to support agroecological transitions in both the South and the North. Its often tacit nature, however, makes it difficult to transmit on a wider scale and via traditional training methods. Based on these observations, the research focused on experimenting with an agroecology training curriculum based on doing, i.e. giving work situations a central place in learning. The experiment was carried out in three West African countries (Benin, Senegal and Togo). An analysis of the results, based on rules of action formulated by the learners at the end of the training and twelve months afterward, reveals that the training provided increased the participants’ power to act. Notably, they put this power to work in the service of the farming community to which they belong. In fact, a significant proportion of the peasants trained act as intermediaries in the circulation of the local knowledge they have learnt, and in this way help to promote the emergence of collective agricultural development projects
Linear B among Indians, Martians and Minoans - or Alice Kober's "science of graphics", Leonard Bloomfield and the American distributionnal linguistics
No full textInternational audienc
Corrigendum to “Co-designing peer-to-peer support in oncology: A participatory study on the development of the PaRole OncO France model” [Patient Educ. Couns. 143 (2026) 109415]
No full textInternational audienc
Interplay of Human and Structural Capital Signals for New Technology Ventures in Early-Stage Venture Capital Funding
No full textInternational audienceBy drawing on the socio-cognitive perspective of signalling theory, this study examines how the interplay of human and structural capital signals may be associated with early-stage venture capital (VC) funding for new technology ventures. Using a dataset of 453 French new technology ventures operating in the digital industry, the findings reveal that strong signals such as patents or trademarks enhance firm legitimacy, while weaker signals such as education level or degrees from elite institutions reinforce team credibility. Both types of signals are positively correlated with early-stage VC funding. However, the results indicate that combining signals of the same dimension and strength yields diminishing returns because additional within-dimension signals are informationally redundant, add little new information about legitimacy or credibility, and lower the likelihood of funding. In noisy, information-saturated contexts, overlapping signals create redundancy and investor cognitive overload rather than facilitating decision-making. This study advances signalling theory by framing signal interactions as a strategic portfolio choice, and outlines implications for entrepreneurial research, investor decision-making, and future work on signalling strategies
