HAL Portal IOGS (nstitut d'Optique Graduate School)
Not a member yet
12589 research outputs found
Sort by
TILES-2018 Sleep Benchmark Dataset: A Longitudinal Wearable Sleep Data Set of Hospital Workers for Modeling and Understanding Sleep Behaviors
No full textsubmitted to IEEE Journal of Biomedical and Health InformaticsSleep is important for everyday functioning, overall well-being, and quality of life. Recent advances in wearable sensing technology have enabled continuous, noninvasive, and cost-effective monitoring of sleep patterns in real-world natural living settings. Wrist-worn devices, in particular, are capable of tracking sleep patterns using accelerometers and heart rate sensors. To support sleep research in naturalistic environments using wearable sensors, we introduce the TILES-2018 Sleep Benchmark dataset, which we make publicly available to the research community. This dataset was collected over a 10-week period from 139 hospital employees and includes over 6,000 unique sleep recordings, alongside self-reported survey data from each participant which include sleep quality, stress, and anxiety among other measurements. We present in-depth analyses of sleep patterns by combining the TILES-2018 Sleep Benchmark dataset with a previously released dataset, (TILES-2018), that follows a similar study protocol. Our analyses include sleep duration, sleep stages, and sleep diaries. Moreover, we report machine learning benchmarks using this dataset as a testbed for tasks including sleep stage classification, prediction of self-reported sleep quality, and classifying demographics. Overall, this dataset provides a valuable resource for advancing foundational studies in sleep behavior modeling
Effect of top metallic contacts on energy conversion performances for near-field thermophotovoltaics
No full textInternational audienceThe design of metallic contact grids on the front side of thermophotovoltaic cells is critical since it can cause significant optical and electrical resistive losses, particularly in the near field. However, from the theoretical point of view, this effect has been either discarded or studied by means of extremely simplified models like the shadowing methods, that consist in simply ignoring the fraction of the semiconductor surface covered by metal. Our study, based on a rigorous three-body theoretical framework and implemented using the scattering matrix approach with the Fourier modal method augmented with adaptive spatial resolution, provides deeper insight into the influence of the front metal contact grid. This approach allows direct access to the radiative power absorbed by the semiconductor, enabling the proposal of an alternative definition for the thermophotovoltaic cell efficiency. By modeling this grid as a metallic grating, we demonstrate its significant impact on the net radiative power absorbed by the cell and, consequently, on the generated electrical power. Our analysis reveals behaviors differing substantially from those predicted by previous simplistic approaches
A Resource-Driven Approach for Implementing CNNs on FPGAs Using Adaptive IPs
No full textInternational audienceThe increasing demand for real-time, low-latency artificial intelligence applications has propelled the use of Field-Programmable Gate Arrays (FPGAs) for Convolutional Neural Network (CNN) implementations. FPGAs offer reconfigurability, energy efficiency, and performance advantages over GPUs, making them suitable for edge devices and embedded systems. This work presents a novel library of resource-efficient convolution IPs designed to automatically adapt to the available FPGA resources. Developed in VHDL, these IPs are parameterizable and utilize fixed-point arithmetic for optimal performance. Four IPs are introduced, each tailored to specific resource constraints, offering flexibility in DSP usage, logic consumption, and precision. Experimental results on a Zynq UltraScale+ FPGA highlight the trade-offs between performance and resource usage. The comparison with recent FPGA-based CNN acceleration techniques emphasizes the versatility and independence of this approach from specific FPGA architectures or technological advancements. Future work will expand the library to include pooling and activation functions, enabling broader applicability and integration into CNN frameworks.</div
Sensor Distance Learning For Cross-Camera Color Constancy
No full textInternational audienceComputational color constancy has seen strong improvement these last years due to the emergence of large labeled datasets. However, the models trained on images acquired by some cameras show low generalization power when tested on images acquired by other cameras. Indeed, since the light chromaticities are device dependent, the training distribution is very spread out when mixing different sensors. In this paper, we propose to inform the network that this complex distribution is a set of simpler distributions, one for each considered camera. For this purpose, we create a Siamese architecture trained with a specific contrastive loss. This loss enforces the model to predict light chromaticities in the same sensor distribution, when considering images acquired by the same camera and in different distributions for images from different sensors. The key idea consists in learning a specific color distance that is sensitive to only sensor variations and not to lighting variations. This learned distance is a nice tool to control if two chromaticity points are in the same sensor distribution or not. We test this original training process in the context of cross-camera color constancy and we show that it outperforms the alternatives on three dataset
Emotion Recognition in Contemporary Dance Performances Using Laban Movement Analysis
No full textInternational audienceThis paper presents a novel framework for emotion recognition in contemporary dance by improving existing Laban Movement Analysis (LMA) feature descriptors and introducing robust, novel descriptors that capture both quantitative and qualitative aspects of the movement. Our approach extracts expressive characteristics from 3D keypoints data of professional dancers performing contemporary dance under various emotional states, and train multiple classifiers, including Random Forests and Support Vector Machines. Additionally, provide in-depth explanation of features and their impact on model predictions using explainable machine learning methods. Overall, our study improves emotion recognition in contemporary dance and offers promising applications in performance analysis, dance training, and human–computer interaction with highest accuracy of 96.85%
Quantum coherence and photon correlations in entangled fluorescent organic molecule pairs
No full textInternational audienc
De l'intrication de quasi-particules dans un condensat de Bose-Einstein
No full textThis thesis focuses on the non-separability of pairs of quasi-particles excited by parametric resonance. The experimental setup used here allows the production of a Bose-Einstein condensate of metastable helium. The use of an ultra-cold atomic gas makes it possible to reach sufficiently low temperatures to observe intrinsically quantum phenomena: the non-separability of the state. In this work, we use the condensate as a coherent reservoir to populate two momentum modes. The advantage of metastable helium is its high internal energy, which allows the electronic detection of single particles. We therefore measure the position and the time of impact of the particles after a time of flight of 308 ms, which allows us to reconstruct the in-trap momentum distribution. In the first theoretical contribution of this work, we demonstrate that measuring the two- and four-body correlation functions not only attests to, but also quantifies the non-separability of a Gaussian state. We also derive a new entanglement witness using only the two-body correlation function. In the experimental part, we improve the machine used to produce our ultra-cold gas and enhance its stability. We implement original techniques to deflect part of the atoms and avoid the saturation of our detector. These improvements allow us to observe the non-separability of the state.Ce mémoire de thèse traite de la non-séparabilité de paires de quasi-particules excitées par résonance paramétrique. Le dispositif expérimental utilisé pendant cette thèse permet de produire un condensat de Bose-Einstein d'hélium métastable. L'utilisation d'un gaz d'atomes ultra-froid permet d'atteindre des températures suffisamment basses afin de pouvoir observer des phénomènes intrinsèquement quantiques : la non-séparabilité de l'état. Dans ce travail, nous utilisons le condensat comme un réservoir cohérent permettant de peupler deux modes d'impulsions. L'avantage de l'hélium métastable est sa grande énergie interne, qui permet la détection électronique de particules uniques. Nous mesurons donc la position et le temps d'impact des particules après un temps de vol de 308 ms, ce qui permet de reconstruire la distribution en impulsion dans le piège. Dans la première contribution théorique de ce travail, nous démontrons que la mesure des fonctions de corrélation à deux et quatre corps permet de quantifier la non-séparabilité d'un état gaussien. Nous dérivons également un critère permettant d'attester la séparabilité de l'état via la seule mesure la fonction de corrélation à deux corps. Dans la partie expérimentale, nous améliorons la machine permettant de produire notre gaz ultra-froid, ainsi que sa stabilité. Par ailleurs, nous mettons en œuvre des techniques originales afin de dévier une partie des atomes et éviter la saturation de notre détecteur. Ces améliorations nous permettent ainsi d'observer la non-séparabilité de l'état
On the Closed-Form of Flow Matching: Generalization Does Not Arise from Target Stochasticity
No full textInternational audienceModern deep generative models can now produce high-quality synthetic samples that are often indistinguishable from real training data. A growing body of research aims to understand why recent methods -such as diffusion and flow matching techniques -generalize so effectively. Among the proposed explanations are the inductive biases of deep learning architectures and the stochastic nature of the conditional flow matching loss. In this work, we rule out the latter -the noisy nature of the loss -as a primary contributor to generalization in flow matching. First, we empirically show that in high-dimensional settings, the stochastic and closed-form versions of the flow matching loss yield nearly equivalent losses. Then, using state-of-the-art flow matching models on standard image datasets, we demonstrate that both variants achieve comparable statistical performance, with the surprising observation that using the closed-form can even improve performance. * Equal contribution
Comportement des nanoparticules sous chauffage par impulsions courtes : Étude des mécanismes atomiques via simulations atomistiques
No full textInternational audienceNanoparticles undergo complex transformations under short pulsed heating, which are crucial for applications across catalysis, material science, and nanotechnology. In this study, we utilize molecular dynamics (MD) simulations to investigate the dynamic behavior of various mono- and bi-metallic nanoparticles and their alloys, under rapid thermal excitation [1]. Employing the ReaxFF reactive force field, we explore processes such as sintering, alloying, and fragmentation in the presence of oxidation and other reactions, providing new insights into the atomic-level mechanisms driving nanoparticle evolution [2]. Our simulations reveal a set of structural and phase transitions governed by pressure-temperature variations. Moreover, nanoparticles are shown to vibrate and when heating exceeds a critical threshold, they experience fragmentation, with smaller particles undergoing more different changes due to surface effects. Additionally, the study highlights the role of nanoparticle size, composition, and thermal conditions, emphasizing how alloying and mixing are enhanced under fast heating regimes. The findings illustrate the potential for controlling nanoparticle properties and morphology by tuning heating rates, particle size, and composition, contributing to the design of advanced nanomaterials for diverse applications, including catalysis, sensors, and energy storage. These results bridge theoretical and experimental efforts, paving the way for optimized synthesis strategies and novel nanoparticle-based materials.References[1]Itina, T. E. (2024). Understanding mono-and bi-metallic Au and Ni nanoparticle responses to fast heating. Nanoscale advances, 6(21), 5451-5463.[2] Itina, T. E., et al. (2025, January). Femtosecond Laser Fabrication of Plasmonic-Magnetic Ni-Au Nanoparticles with Enhanced Magnetic Properties. Photonics West, LASE, Conference 13352 Nanoscale and Quantum Materials: From Synthesis and Laser Processing to Applications 2025 25-26 January 2025, San Francisco, USA.Les nanoparticules subissent des transformations complexes sous l'effet de courts chauffages pulsés, des transformations cruciales pour des applications en catalyse, science des matériaux et nanotechnologie. Dans cette étude, nous utilisons des simulations de dynamique moléculaire (DM) pour étudier le comportement dynamique de diverses nanoparticules mono- et bimétalliques et de leurs alliages, sous excitation thermique rapide[1]. Grâce au champ de force réactif ReaxFF, nous explorons des processus tels que le frittage, l'alliage et la fragmentation en présence d'oxydation et d'autres réactions, apportant ainsi de nouvelles perspectives sur les mécanismes à l'échelle atomique qui régissent l'évolution des nanoparticules [2]. Nos simulations révèlent un ensemble de transitions structurelles et de phases régies par les variations de pression et de température. De plus, les nanoparticules vibrent et, lorsque le chauffage dépasse un seuil critique, se fragmentent, les particules plus petites subissant des modifications plus variées dues aux effets de surface. De plus, l'étude met en évidence le rôle de la taille, de la composition et des conditions thermiques des nanoparticules, en insistant sur l'amélioration de l'alliage et du mélange sous des régimes de chauffage rapide. Ces résultats illustrent le potentiel de contrôle des propriétés et de la morphologie des nanoparticules en ajustant les vitesses de chauffage, la taille des particules et leur composition, contribuant ainsi à la conception de nanomatériaux avancés pour diverses applications, notamment la catalyse, les capteurs et le stockage d'énergie. Ces résultats relient les efforts théoriques et expérimentaux, ouvrant la voie à des stratégies de synthèse optimisées et à de nouveaux matériaux à base de nanoparticules.References[1]Itina, T. E. (2024). Understanding mono-and bi-metallic Au and Ni nanoparticle responses to fast heating. Nanoscale advances, 6(21), 5451-5463.[2] Itina, T. E., et al. (2025, January). Femtosecond Laser Fabrication of Plasmonic-Magnetic Ni-Au Nanoparticles with Enhanced Magnetic Properties. Photonics West, LASE, Conference 13352 Nanoscale and Quantum Materials: From Synthesis and Laser Processing to Applications 2025 25-26 January 2025, San Francisco, USA
Pattern-Based Graph Classification: Comparison of Quality Measures and Importance of Preprocessing
No full textInternational audienceGraph classification aims to categorize graphs based on their structural and attribute features, with applications in diverse fields such as social network analysis and bioinformatics. Among the methods proposed to solve this task, those relying on patterns (i.e. subgraphs) provide good explainability, as the patterns used for classification can be directly interpreted. To identify meaningful patterns, a standard approach is to use a quality measure, i.e. a function that evaluates the discriminative power of each pattern. However, the literature provides tens of such measures, making it difficult to select the most appropriate for a given application. Only a handful of surveys try to provide some insight by comparing these measures, and none of them specifically focuses on graphs. This typically results in the systematic use of the most widespread measures, without thorough evaluation. To address this issue, we present a comparative analysis of 38 quality measures from the literature. We characterize them theoretically, based on four mathematical properties. We leverage publicly available datasets to constitute a benchmark, and propose a method to elaborate a gold standard ranking of the patterns. We exploit these resources to perform an empirical comparison of the measures, both in terms of pattern ranking and classification performance. Moreover, we propose a clustering-based preprocessing step, which groups patterns appearing in the same graphs to enhance classification performance. Our experimental results demonstrate the effectiveness of this step, reducing the number of patterns to be processed while achieving comparable performance. Additionally, we show that some popular measures widely used in the literature are not associated with the best results