INRIA a CCSD electronic archive server
Not a member yet
122212 research outputs found
Sort by
Leveraging Data Seasonality and Matrix Profile for Anomaly Detection: Application to Climate Time Series
No full textSeasonal time series analysis is fundamental in domains such as climate science, where detecting and understanding anomalies, patterns, and data changes are essential. The classical Matrix Profile approach does not consider the data's seasonality, failing to detect seasonal anomalies and patterns. This paper introduces the Interval Matrix Profile, a novel extension of the Matrix Profile specifically designed for analyzing periodic and seasonal time series data. The Interval Matrix Profile enables flexible interval-based comparisons across seasons, allowing the detection of anomalies that conventional approaches miss. We further propose the constrained k Nearest Neighbor Interval Matrix Profile, designed to identify anomalies that may appear across multiple periods, a common characteristic of abnormal climate events and extreme weather phenomena. Our approach leverages a scalable block-based algorithm that achieves significant performance gains through caching, vectorization, and parallelism. Additionally, we introduce a novel methodology to detect the first or last occurrence of a pattern, enabling the discovery of pattern emergence or disappearance within seasonal time series. The algorithms are demonstrated in case studies on temperature climate time series. They effectively capture seasonal anomalies and find pattern disappearance. Our results illustrate that the IMP consistently outperforms the classical Matrix Profile in the accuracy of seasonal anomaly detection and computational efficiency
gGRAPPA: A Flexible, GPU-Accelerated Python Package for Fast and Efficient generalized GRAPPA Reconstruction
No full textInternational audienceMotivation: Existing open-source MR reconstruction tools often fail to efficiently utilize GPU resources and lack support for generalized GRAPPA implementations. Many tools are limited to 2D or 3D reconstruction, and few incorporate advanced techniques such as 2D-CAIPIRINHA, which enhances imaging capabilities. Goal(s): gGRAPPA aims to provide a fast, flexible, and open-source tool for generalized GRAPPA/CAIPI reconstruction. Approach: By utilizing PyTorch, gGRAPPA runs multiple convolutional windows in batch mode to optimize GPU memory usage and accelerate reconstruction times. Results: gGRAPPA achieves up to a 65x speedup over CPU implementations and a 6x speedup compared to non-batched GPU methods, enabling efficient and fast reconstruction MRI scans.Purpose: Generalized autocalibrating partially parallel acquisitions (GRAPPA) is widely used in MRI reconstruction, but conventional implementations, such as those in MATLAB on CPUs, are computationally intensive, especially for high-dimensional data, and those in Python on GPU not implementing a "generalized GRAPPA". We developed gGRAPPA to address these limitations by (1) utilizing GPU acceleration for high-speed processing and (2) supporting flexible GRAPPA configurations (2D and 3D). Furthermore, gGRAPPA allows users to have full control on the reconstruction by tuning key parameters like kernel size and regularization strength. This tool aims to provide a generalized and flexible GRAPPA reconstruction framework allowing fast and efficient processing on GPU.Impact: gGRAPPA provides a fast, flexible, and open-source solution for GRAPPA MRI reconstruction on GPU, significantly accelerating reconstruction times and enabling ultra high-resolution imaging reconstruction, thereby supporting advanced research applications across diverse MRI protocols.</div
Réseau de neurones convolutifs sur graphes multimodal pour l'étude de la structure et de la fonction cérébrales dans l'anxiété et la dépression à l'adolescence
No full textNational audienceL’anxiété et la dépression sont des troubles mentaux affectant négativement la qualité de vie. Plusieurs études ont développé des approches multimodales à partir de données issues de l’Imagerie par Résonance Magnétique (IRM) pour exploiter des informations complémentaires à travers plusieurs modalités, afin de mieux comprendre les mécanismes biologiques derrière ces maladies. Récemment, les réseaux de neurones convolutifs sur graphes (GCN) ont émergé comme un outil puissant dans la recherche sur la connectivité cérébrale. Ici, nous développons un modèle GCN multimodal pour modéliser conjointement la structure et la fonction cérébrales afin de classifier l’anxiété et la dépression chez les adolescents, en utilisant la base de données Boston Adolescent Neuroimaging of Depression and Anxiety (BANDA).La topologie du graphe est initialisée à partir de la connectivité structurelle dérivée de l’IRM de diffusion, tandis que la connectivité fonctionnelle est intégrée sous forme de features sur les noeuds pour améliorer la distinction entre les patients anxieux, dépressifs et les controls sains. L’interprétation des principales régions cérébrales contribuant à la classification est réalisée grâce à la méthode Gradient-weighted Class Activation Mapping (Grad-CAM). Des approches unimodales utilisant exclusivement des données structurelles ou fonctionnelles ont aussi été étudiées pour établir des tests comparatifs entre notre modèle GCN multimodal et ceux unimodaux. Les performances obtenues sont supérieures dans la majorité des tâches de classification, révélant la puissance discriminative de notre approche multimodale, et mettent en évidence des motifs significatifs d’altérations cérébrales associées à l’anxiété et à la dépression
Joint Learning of Linear Dynamical Systems under Smoothness Constraints
No full textWe consider the problem of joint learning of multiple linear dynamical systems. This has received significant attention recently under different types of assumptions on the model parameters. The setting we consider involves a collection of linear systems each of which resides on a node of a given undirected graph . We assume that the system matrices are marginally stable, and satisfy a smoothness constraint w.r.t -- akin to the quadratic variation of a signal on a graph. Given access to the states of the nodes over time points, we then propose two estimators for joint estimation of the system matrices, along with non-asymptotic error bounds on the mean-squared error (MSE). In particular, we show conditions under which the MSE converges to zero as increases, typically polynomially fast w.r.t . The results hold under mild (i.e., ), or sometimes, even no assumption on (i.e. )
A Packet Collision Avoidance Resource Selection Scheme for Reliable Intra-Platoon Message Delivery in a C-V2X network
No full textInternational audienceReliable intra-platoon communication is critical for safety-related message delivery within a platoon of connected and automated vehicles. However, the intra-platoon communication is challenged by packet collisions due to hidden nodes and merging collisions due to vehicle mobility. To address these challenges, this paper proposes a packet collision avoidance resource selection (PCA-RS) scheme to enhance the standardized SPS scheme. The proposed PCA-RS scheme introduces three enhancement mechanisms, which aims to alleviate merging collisions and hidden-node collisions in intra-platoon message delivery. A resource partition mechanism is introduced to divide frequency-time resources in a selection window into two sets in order for vehicles (both platoon and non-platoon vehicles) moving in opposite directions to select different frequency-time resources and thus avoid potential merging collisions; an intra-platoon cooperative mechanism is introduced to enable the leader of a platoon to know the resource occupation status on the hidden nodes of the platoon according to the messages received from the last platoon member of the same platoon and thus avoid potential hidden-node collisions; and a merging collision detection mechanism is introduced to enable a non-platoon vehicle to detect the status of the frequency-time resources it currently occupies after the non-platoon vehicle changes a lane and thus avoids potential merging collisions among non-platoon vehicles due to lane-change maneuvers. Simulation results demonstrate that compared with the standardized SPS scheme, the proposed PCA-RS scheme can improve the reliability of intra-platoon message delivery in terms of the intra-platoon packet delivery ratio
Movements in collaborative tools. Evolutionary dynamics if civil security's artifact ecologies.
No full textInternational audiencePoster presented at the PePR eNSEMBLE annual days in january 2025. It summarizes the first results of our PhD Thesis research surrounding pre-existing artifact ecologies in Frenchs SDIS and 7 identifyed dynamic movements inside those artifact ecologies
Forbidden Patterns in Temporal Graphs Resulting from Encounters in a Corridor
No full textInternational audienceIn this paper, we study temporal graphs arising from mobility models, where verticescorrespond to agents moving in space and edges appear each time two agents meet. Wepropose a rather natural one-dimensional model.If each pair of agents meets exactly once, we get a simple temporal clique where theedges are ordered according to meeting times. In order to characterize which temporal cliquescan be obtained as such ‘mobility graphs’, we introduce the notion of forbidden patterns intemporal graphs. Furthermore, using a classical result in combinatorics, we count the numberof such mobility cliques for a given number of agents, and show that not every temporalclique resulting from the 1D model can be realized with agents moving with different constantspeeds. For the analogous circular problem, where agents are moving along a circle, we providea characterization via circular forbidden patterns.Our characterization in terms of forbidden patterns can be extended to the case whereeach edge appears at most once. We also study the problem where pairs of agents are allowedto cross each other several times, using an approach from automata theory. We observe thatin this case, there is no finite set of forbidden patterns that characterize such temporal graphsand nevertheless give a linear-time algorithm to recognize temporal graphs arising from thismodel
Efficient Hamiltonian, structure and trace distance learning of Gaussian states
No full textIn this work, we initiate the study of Hamiltonian learning for positive temperature bosonic Gaussian states, the quantum generalization of the widely studied problem of learning Gaussian graphical models. We obtain efficient protocols, both in sample and computational complexity, for the task of inferring the parameters of their underlying quadratic Hamiltonian under the assumption of bounded temperature, squeezing, displacement and maximal degree of the interaction graph. Our protocol only requires heterodyne measurements, which are often experimentally feasible, and has a sample complexity that scales logarithmically with the number of modes. Furthermore, we show that it is possible to learn the underlying interaction graph in a similar setting and sample complexity. Taken together, our results put the status of the quantum Hamiltonian learning problem for continuous variable systems in a much more advanced state when compared to spins, where state-of-the-art results are either unavailable or quantitatively inferior to ours. In addition, we use our techniques to obtain the first results on learning Gaussian states in trace distance with a quadratic scaling in precision and polynomial in the number of modes, albeit imposing certain restrictions on the Gaussian states. Our main technical innovations are several continuity bounds for the covariance and Hamiltonian matrix of a Gaussian state, which are of independent interest, combined with what we call the local inversion technique. In essence, the local inversion technique allows us to reliably infer the Hamiltonian of a Gaussian state by only estimating in parallel submatrices of the covariance matrix whose size scales with the desired precision, but not the number of modes. This way we bypass the need to obtain precise global estimates of the covariance matrix, controlling the sample complexity
Brain network alignment using structural and functional connectivity with anatomical constraints
No full textBrain research often segments the cortex into regions based on an atlas, assuming a perfect matching of the regions across different subjects. However, due to genetic and environmental factors, inter-subject variability makes it challenging to produce a single brain atlas with a perfect correspondence of region labels across subjects, especially at a fine-grained scale. Previous work has proposed to use structural connectomes constructed from diffusion magnetic resonance imaging to permute brain regions across subjects and align the cortical regions from one subject to another, leading to an improved similarity of connectomes across subjects. In this work, we propose a multimodal approach to exploit simultaneously structural and functional connectivity information in the alignment process. Spatial constraints are included to prevent unlikely permutations between remote regions. Experimental results show the validity of the approach and the effectiveness of the constraint
A systematic review of immersive technologies for education: effects of cognitive load and curiosity state on learning performance
No full textInternational audienceImmersive technologies are assumed to have many benefits for learning due to their potential positive impact on optimizing learners' cognitive load and fostering intrinsic motivation. However, despite promising results, the findings regarding the actual impact on learning remain inconclusive, raising questions about the determinants of efficacy. To address these gaps, we conducted a PRISMA systematic review to investigate the contributions and limitations of Virtual Reality (VR) and Augmented Reality (AR) in learning, specifically by examining their effects on cognitive load and intrinsic motivations. Through the application of an analytical grid, we systematically classified the impact of VR/AR on the causal relationship between learning performance (i.e., objective learning improvement) and cognitive load or motivation, while respecting the fundamental assumptions of the main theories related to these factors.Analyzing 36 studies, the findings reveal that VR, often causing extraneous load, hinders learning, particularly among novices. In contrast, AR optimizes cognitive load, proving beneficial for novice learners but demonstrating less effectiveness for intermediate learners. The effects on intrinsic motivation remain inconclusive, likely due to variations in measurement methods. The review underscores the need for detailed, sophisticated evaluations and comprehensive frameworks that consider both cognitive load and intrinsic motivation to improve understanding of the impact of immersive technologies on learning