INRIA a CCSD electronic archive server
Not a member yet
122212 research outputs found
Sort by
HOT-POT: Optimal Transport for Sparse Stereo Matching
No full textStereo vision between images faces a range of challenges, including occlusions, motion, and camera distortions, across applications in autonomous driving, robotics, and face analysis. Due to parameter sensitivity, further complications arise for stereo matching with sparse features, such as facial landmarks. To overcome this ill-posedness and enable unsupervised sparse matching, we consider line constraints of the camera geometry from an optimal transport (OT) viewpoint. Formulating camera-projected points as (half)lines, we propose the use of the classical epipolar distance as well as a 3D ray distance to quantify matching quality. Employing these distances as a cost function of a (partial) OT problem, we arrive at efficiently solvable assignment problems. Moreover, we extend our approach to unsupervised object matching by formulating it as a hierarchical OT problem. The resulting algorithms allow for efficient feature and object matching, as demonstrated in our numerical experiments. Here, we focus on applications in facial analysis, where we aim to match distinct landmarking conventions.18 pages, 10 figures, 6 table
Quasiprobabilistic imaginary-time evolution on quantum computers
No full textImaginary-time evolution plays an important role in algorithms for computing ground-state and thermal equilibrium properties of quantum systems, but can be challenging to simulate on classical computers. Many quantum algorithms for imaginary-time evolution have resource requirements that are prohibitive for current quantum devices and face performance issues due to noise. Here, we propose a new algorithm for computing imaginary-time evolved expectation values on quantum computers, inspired by probabilistic error cancellation, an error-mitigation technique. Our algorithm works by decomposing a Trotterization of imaginary-time evolution into a probabilistic linear combination of operations, each of which is then implemented on a quantum computer. The measurement data is then classically post-processed to obtain the expectation value of the imaginary-time evolved state. Our algorithm requires no ancillary qubits and can be made noise-resilient without additional error mitigation. It is well-suited for estimating thermal expectation values by making use of the notion of a thermal pure quantum state. We demonstrate our algorithm by performing numerical simulations of thermal pure quantum state preparation for the 1D Heisenberg Hamiltonian on 8 qubits, and by using an IBM quantum computer to estimate the energy of the same Hamiltonian on 2 qubits. We observe promising results compared to the exact values, illustrating the potential of our algorithm for probing the physics of quantum many-body systems on current hardware
The SChISM study: Circulating cell-free DNA size profiles as predictors of progression in advanced carcinoma treated with immune-checkpoint inhibitors
No full textBackgroundCirculating cell-free DNA (cfDNA) offers a promising noninvasive way to predict resistance to immune-checkpoint inhibitors (ICI), for which robust biomarkers are still lacking.MethodsThe SChISM (Size CfDNA Immunotherapy Signature Monitoring) proof-of-concept study (NCT05083494) collected baseline plasmatic cfDNA size profiles from 126 ICI-treated advanced carcinomas, quantified using the innovative, patented and standardized BIABooster device (Adelis). Fragmentome-derived variables and standard clinical variables (including neutrophils-to-lymphocyte ratio, NLR) were analyzed for univariable associations with early progression (EP, progression at first imaging) and progression-free survival (PFS). Multivariable analysis was carried through both unsupervised and supervised learning. Twenty-six variable selection methods combined with 11 models were benchmarked to derive a multivariable predictive model relying on a minimal subset of variables.ResultsHigher cfDNA concentration and high quantities of short fragments (111-240 base pairs (bp)) were associated with poor response and reduced PFS, unlike long fragments (> 300 bp). The proportion of fragments longer than 1650 bp exhibited the strongest association, with non-EP odds ratio = 0.39 [95% CI: 0.25-0.62] and PFS hazard ratio = 0.54 [95% CI: 0.42-0.68]. Unsupervised learning identified four patient clusters significantly associated with EP (p=0.004, Pearson’s Chi-squared test) and PFS (p=0.001, log-rank test). The multivariable machine learning analysis identified a subset of nine variables that shown greater performances in a logistic regression model: EP positive predictive value compared to single marker AUC_{R_{[>1650]}}=73.6±3.70\% PPV_{R_{[>1650]}}=55.8±7.46\% .ConclusioncfDNA size profiles significantly associated with progression and PFS during ICI, outperforming the routinely used markers.</p
Detecting and Measuring Client-and Server-Side Google Tag Manager and its Tags in 80K Websites
No full textInternational audienceGoogle Tag Manager (GTM) enables website Publishers to install analytics, advertising, and other tracking services as "Tags" on their websites. Traditionally, in Client-Side GTM, Tags collect and send data directly from the browser to third-party servers. However, as today's browsers and extensions increasingly block third-party trackers, Google introduced a Server-Side GTM in 2020. This GTM version allows to install and execute Tags directly on a first-party server, thus hiding the presence of third-party trackers. In this study, we perform a crawl of 80K popular websites worldwide to analyze the adoption of Client-and Server-Side GTM and the prevalence of their Tags. Our results show that GTM is present on 28.8% of websites, with 6.7% of these sites obfuscating its presence. Our Tags detection methodology is able to measure the prevalence of all GTM Tags -Official, Template, HTML and Image Tags -globally and reveals a strong dominance of Tags provided by Google, which appear on 95.3% of sites. We further analyze 386K HTML Tags and find that they invoke third-party libraries on at least 81.2% of websites, to potentially share personal data. Finally, we detect the presence of Server-Side GTM on 3K websites and 398 Server-Side Tags instances on the Web. We also propose a GTM-Eye browser extension that detects GTM and its Tags accessible to everyone
Collaborative Action on Timing InterferenCes: Summary and Perspectives at Mid-term
No full textTo appear in Embedded Real Time Systems (ERTS) 2026International audienceCAOTIC is an ambitious initiative aimed at pooling and coordinating the efforts of major French research teams working on timing analysis of multicore real-time systems, with a focus on interference due to shared resources. The objective is to enable the efficient use of multicores in critical systems. Based on a better understanding of timing anomalies and interference, considering the specificities of applications (structural properties and execution model), and revisiting the links between timing analysis and synthesis processes (code generation, mapping, scheduling), we target significant progresses in timing analysis models and techniques for critical systems, as well as in methodologies for their application in industry. In this paper, at project mid-term, we show the progress of the project. We also present some original work, about the use of a Tricore plaform and its timing model, and discuss open questions and future work
The Third VoicePrivacy Challenge: Preserving Emotional Expressiveness and Linguistic Content in Voice Anonymization
No full textWe present results and analyses from the third VoicePrivacy Challenge held in 2024, which focuses on advancing voice anonymization technologies. The task was to develop a voice anonymization system for speech data that conceals a speaker's voice identity while preserving linguistic content and emotional state. We provide a systematic overview of the challenge framework, including detailed descriptions of the anonymization task and datasets used for both system development and evaluation. We outline the attack model and objective evaluation metrics for assessing privacy protection (concealing speaker voice identity) and utility (content and emotional state preservation). We describe six baseline anonymization systems and summarize the innovative approaches developed by challenge participants. Finally, we provide key insights and observations to guide the design of future VoicePrivacy challenges and identify promising directions for voice anonymization research.</div
The Branch-and-Bound Tree Closure
No full textThis paper investigates the a-posteriori analysis of Branch-and-Bound (BB) trees to extract structural information about the feasible region of mixed-binary linear programs. We introduce three novel outer approximations of the feasible region, systematically constructed from a BB tree. These are: a tight formulation based on disjunctive programming, a branching-based formulation derived from the tree's branching logic, and a mixing-set formulation derived from the on-off properties inside the tree. We establish an inclusion hierarchy, which ranks the approximations by their theoretical strength w.r.t. to the original feasible region. The analysis is extended to the generation of valid inequalities, revealing a separation-time hierarchy that mirrors the inclusion hierarchy in reverse. This highlights a trade-off between the tightness of an approximation and the computational cost of generating cuts from it. Motivated by the computational expense of the stronger approximations, we introduce a new family of valid inequalities called star tree inequalities. Although their closure forms the weakest of the proposed approximations, their practical appeal lies in an efficient, polynomial-time combinatorial separation algorithm. A computational study on multi-dimensional knapsack and set-covering problems empirically validates the theoretical findings. Moreover, these experiments confirm that computationally useful valid inequalities can be generated from BB trees obtained by solving optimization problems considered in practice
The Radial Spanning Tree is straight in all dimensions
No full textThe Radial Spanning Tree (RST) in dimension is a random geometric graph constructed on a homogeneous Poisson point process in augmented by the origin, with edges connecting each to the nearest point that lies closer to than , with respect to the Euclidean distance. By construction, it forms almost surely a tree rooted at . The RST was introduced in 2007 by Baccelli and Bordenave, who investigated straightness, a deterministic property introduced by Howard and Newman in 2001, to derive information about the asymptotic directions of infinite branches. They proved that the RST is almost surely straight in dimension , which directly implies that all infinite branches are asymptotically directed, every possibility is attained, and directions reached by multiple infinite branches form a dense subset. However, their approach relies crucially on planarity, preventing any straightforward extension to higher dimensions. In this paper, we close this gap by proving that the RST is almost surely straight in any dimension, thereby obtaining the same consequences for the behaviour of infinite branches. Our approach resolves the key barriers in the study of the RST, notably those posed by the complex dependency structure combined with the radial nature of the model, and especially beyond the planar setting. It relies on tools developed for the analysis of the Directed Spanning Forest, a closely related model, including recent progress by the author in 2025. Specifically, a key contribution of this work is the construction of a suitable renewal-type decomposition of RST paths. Leveraging this decomposition together with classical concentration inequalities, we show that RST paths cannot deviate far from straight lines and derive straightness
LivingBench: an IoT/Edge Platform Benchmark Based on an Environmental Observation Use-Case
No full textInternational audienceIn recent years, a number of edge computing platforms have been proposed to process data produced by IoT sensors. Performing computation close to the sources of data allows faster insight and greater reliability at a lower cost compared to traditional cloud-based deployments. However, designers of IoT/edge platforms face difficult issues. In particular, exercising and testing a new platform in conditions that approach a real deployment requires a sufficient number of standard benchmarking systems capable of generating realistic workloads. In this paper, we propose LivingBench, a benchmarking tool with the capability of incorporating real or synthetic workload injection, developed to exercise edge computing systems. LivingBench integrates a real-world data trace captured in an environmental observatory, together with a collection of actual applications designed for processing these data, and a load injector tool capable of replaying a (possibly pre-processed) trace to benchmark an MQTT-based edge system. We describe the architecture of LivingBench and show how it may be used to evaluate the maximum data processing capacity of an edge system under test
Tubes in sub-Riemannian geometry and a Weyl's invariance result for curves in the Heisenberg groups
No full textThe purpose of the paper is threefold: first, we prove optimal regularity results forthe distance from Ck submanifolds of general rank-varying sub-Riemannian structures. Then,we study the asymptotics of the volume of tubular neighbourhoods around such submanifolds.Finally, for the case of curves in the Heisenberg groups, we prove a Weyl’s invariance result:the volume of small tubes around curves does not depend on the way the curve is isometricallyembedded, but only on its Reeb angle. The proof does not need the computation of the actualvolume of the tube, and it is new even for the three-dimensional Heisenberg group