Portail HAL Ensta
Not a member yet
11080 research outputs found
Sort by
A Medical Low-Back Pain Physical Rehabilitation Dataset for Human Body Movement Analysis
No full textInternational audienceWhile automatic monitoring and coaching of exercises are showing encouraging results in non-medical applications, they still have limitations such as errors and limited use contexts. To allow the development and assessment of physical rehabilitation by an intelligent tutoring system, we identify in this article four challenges to address and propose a medical dataset of clinical patients carrying out low back-pain rehabilitation exercises. The dataset includes 3D Kinect skeleton positions and orientations, RGB videos, 2D skeleton data, and medical annotations to assess the correctness, and error classification and localisation of body part and timespan. Along this dataset, we perform a complete research path, from data collection to processing, and finally a small benchmark. We evaluated on the dataset two baseline movement recognition algorithms, pertaining to two different approaches: the probabilistic approach with a Gaussian Mixture Model (GMM), and the deep learning approach with a Long-Short Term Memory (LSTM). This dataset is valuable because it includes rehabilitation relevant motions in a clinical setting with patients in their rehabilitation program, using a cost-effective, portable, and convenient sensor, and because it shows the potential for improvement on these challenges
StickToYourRoleLeaderboard
No full texthttps://huggingface.co/spaces/flowers-team/StickToYourRoleLeaderboardThe Stick to Your Role! leaderboard compares LLMs based on undesired sensitivity to context change. LLM-exhibited behavior always depends on the context (prompt). While some context-dependence is desired (e.g. following instructions), some is undesired (e.g. drastically changing the simulated value expression based on the interlocutor). As proposed in our paper, undesired context-dependence should be seen as a property of LLMs - a dimension of LLM comparison (alongside others such as model size speed or expressed knowledge). This leaderboard aims to provide such a comparison and extends our paper with a more focused and elaborate experimental setup. Standard benchmarks present many questions from the same minimal contexts (e.g. multiple choice questions), we present the same questions from many different contexts
Ultrafast Electron Dynamics in Coupled and Uncoupled HgTe Quantum Dots
No full textInternational audienceIn this article, we study electron dynamics in HgTe quantum dots with a 1.9 μm gap, a material relevant for infrared sensing and emission, using ultrafast spectroscopy with 35 fs time resolution. Experiments have been carried out at several probing photon energies around the gap, which allows us to follow the relaxation path of the photoexcited electrons. We compare such dynamics in two kind of samples, HgTe quantum dots with long ligands and with short ligands, in order to distinguish the role of the coupling between adjacent quantum dots. Three main dynamics can be observed in the transient reflectivity on both samples, with slightly different relaxation times: two fast decays on the time scale of hundreds of femtoseconds and a few picoseconds, respectively, followed by a slower relaxation back to the unperturbed value over hundreds of picoseconds. The two fast components are associated with intraband relaxation of the photoexcited electrons within the conduction band, while the final relaxation path can be assigned to Auger relaxation mechanisms and to the slower interband exciton recombination
Variability Management in Self-Adaptive Systems through Deep Learning: A Dynamic Software Product Line Approach
No full textInternational audienceSelf-adaptive systems can autonomously adjust their behavior in response to environmental changes. Nowadays, not only can these systems be engineered individually, but they can also be conceived as members of a family based on the approach of dynamic software product lines. Through systematic mapping, we build on the identified gaps in variability management of self-adaptive systems, propose a framework that improves the adaptive capability of self-adaptive systems through feature model generation, variation point generation, selection of a variation point, and runtime variability management using deep learning and monitor -analysis -plan -execute -knowledge (MAPE-K) control loop. We compute the permutation of domain features and obtain all the possible variation points that a feature model can possess. After identifying variation points, we obtain an adaptation rule for each variation point of the corresponding product line through a two-stage training of an artificial neural network. To evaluate our proposal, we developed a test case in the context of an air quality-based activity recommender system, in which we generated 11 features and 32 possible variations. The results obtained with the proof of concept show that it is possible to manage identifying new variation points at runtime using deep learning. Future research will employ generating and building variation points using artificial intelligence techniques.</div
A Definition of Open-Ended Learning Problems for Goal-Conditioned Agents
No full textA lot of recent machine learning research papers have ``open-ended learning'' in their title. But very few of them attempt to define what they mean when using the term. Even worse, when looking more closely there seems to be no consensus on what distinguishes open-ended learning from related concepts such as continual learning, lifelong learning or autotelic learning. In this paper, we contribute to fixing this situation. After illustrating the genealogy of the concept and more recent perspectives about what it truly means, we outline that open-ended learning is generally conceived as a composite notion encompassing a set of diverse properties. In contrast with previous approaches, we propose to isolate a key elementary property of open-ended processes, which is to produce elements from time to time (e.g., observations, options, reward functions, and goals), over an infinite horizon, that are considered novel from an observer's perspective. From there, we build the notion of open-ended learning problems and focus in particular on the subset of open-ended goal-conditioned reinforcement learning problems in which agents can learn a growing repertoire of goal-driven skills. Finally, we highlight the work that remains to be performed to fill the gap between our elementary definition and the more involved notions of open-ended learning that developmental AI researchers may have in mind
The Codac Library
No full textInternational audienceCodac (Catalog Of Domains And Contractors) is a C++/Python library providing tools for constraint programming over reals, trajectories and sets. It has many applications in parameter estimation, guaranteed integration or robot localization and provides reliable outputs by computing sets of feasible solutions according to the constraints defining the problem. This paper provides a brief overview of the library and its Contractor Network approach, illustrated on a convincing robotic application
Perceptual evaluation of sound synthesis of wind turbine noise
No full textInternational audienceThis study brings together several laboratories with the aim of assessing the health effects of "audible" noise (> 20 Hz) and infrasound (<20 Hz) emitted by wind turbines. To study loudness and annoyance due to this noise, perceptual tests are planned at the LMA, where a restitution cabin has been developed, specifically designed to diffuse very low frequencies and infrasound. As recording wind turbine noise is only possible at low wind speeds for a good quality sound reproduction, it would be interesting to be able to use sound synthesis of wind turbine noise. From sounds recorded in a wind farm for different meteorological conditions, the corresponding sounds have been synthesized. A physical model synthesis was performed, based on an extended-source aeroacoustic model taking into account propagation over flat ground. Dissimilarity tests including recorded and synthesized sounds enabled a 2D perceptual space to be built. Synthesized and the corresponding recorded sounds are closed together in the perceptual space, but some differences can be perceived, mainly due to difference in amplitude of fluctuation and spectral balance. The analysis of the perceptual space opens up interesting prospects for improving the sound synthesis and its use for future perceptual tests
Constrained and Vanishing Expressivity of Quantum Fourier Models
No full textIn this work, we highlight an unforeseen behavior of the expressivity of Parameterized Quantum Circuits (PQC) for machine learning. A large class of these models, seen as Fourier Series which frequencies are derived from the encoding gates, were thought to have their Fourier coefficients mostly determined by the trainable gates. Here, we demonstrate a new correlation between the Fourier coefficients of the quantum model and its encoding gates. In addition, we display a phenomenon of vanishing expressivity in certain settings, where some Fourier coefficients vanish exponentially when the number of qubits grows. These two behaviors imply novel forms of constraints which limit the expressivity of PQCs, and therefore imply a new inductive bias for Quantum models. The key concept in this work is the notion of a frequency redundancy in the Fourier series spectrum, which determines its importance. Those theoretical behaviours are observed in numerical simulations
Validation par intervalles d'un estimateur non linéaire
No full textInternational audienceIn engineering, models are often used to represent the behavior of a system. Estimators are then needed to approximate the values of the model's parameters based on observations. This approximation implies a difference between the values predicted by the model and the observations that have been made. It creates an uncertainty that can lead to dangerous decision making. Interval analysis tools can be used to guarantee some properties of an estimator, even when the estimator itself doesn't rely on interval analysis (Adam, 2019) (Adam, 2015). This paper contributes to this dynamic by proposing an interval-based and guaranteed method to validate a nonlinear estimator. It is based on the Moore-Skelboe algorithm (van Emden, 2004). This method returns a guaranteed maximum error that the estimator will never exceed. We will show that we can guarantee properties even when working with non-guaranteed estimators such as neural networks
Fast and accurate boundary integral equation methods for the multi-layer transmission problem
No full textInternational audienceWe consider a multi-layer transmission problem, which can be used for example to describe the light scattering in meta-materials (assemblings of various concentric penetrable materials). Our goal is to solve the multi-layer problem accurately with optimal discretization. Generally, the costs to solve this problem grow as more layers are introduced - solving this problem is thus particularly challenging for 3D models. Forthis reason, we use boundary integral equation (BIE) methods: they reduce the dimensionality of the problem and can provide high order accuracy. However, BIE methods suffer from the so-called close evaluation problem. We address it using modified representations. We further examine how to improve the speed of ourmethod by optimizing the accuracy over number of discretization points ratio. In particular, we investigate whether the usual rule of thumb to mesh interfaces, based on the most constraining material, is necessary for the multi-layer transmission problem