École Polytechnique Fédérale de Lausanne
Infoscience - École polytechnique fédérale de LausanneNot a member yet
191401 research outputs found
Sort by
Measurement of the charm mixing parameter y(CP)-y(CP)(K pi) using two-body D-0 meson decays
No full textA measurement of the ratios of the effective decay widths of D-0 -> pi(-)pi(+) and D-0 -> K- K+ decays over that of D-0 -> K-pi(+) decays is performed with the LHCb experiment using proton-proton collisions at a centre-of-mass energy of 13 TeV, corresponding to an integrated luminosity of 6 fb(-1). These observables give access to the charm mixing parameters y(CP)(pi pi) - y(CP)(K pi )and y(CP)(KK) -y(CP)(K pi), and are measured as y(CP)(pi pi) - y(CP)(K pi) = (6.57 +/- 0.53 +/- 0.16) x 10(-3), y(CP)(KK) - y(CP)(K pi) = (7.08 +/- 0.30 +/- 0.14) x 10(-3), where the first uncertainties are statistical and the second systematic. The combination of the two measurements is Y-CP - y(CP)(K pi) = (6.96 +/- 0.26 +/- 0.13) x 10(-3), which is four times more precise than the previous world average.LPH
A Blueprint for Integrating Task-Oriented Conversational Agents in Education
No full textOver the past few years, there has been an increase in the use of chatbots for educational purposes. Nevertheless, the chatbot technologies and architectures that are often applied to educational contexts are not necessarily designed for such contexts. While general-purpose chatbot technologies can be used in educational contexts, there are some challenges specific to these contexts that need to be taken into consideration. Namely, chatbot technologies intended for education should, by design, integrate directly within online learning applications and focus on achieving learning goals by supporting learners with the task at hand. In this paper, we propose a blueprint for an architecture specifically aimed at integrating task-oriented chatbots to support learners in educational contexts. We then present a proof-of-concept implementation of our blueprint as a part of a code review application designed to teach programming best practices. Our blueprint could serve as a starting point for developers in education looking to build chatbot technologies targeting educational contexts and is a first step toward an open chatbot architecture explicitly tailored for learning applications.SCI-STI-DGAVP-E-LEAR
Searching for visual patterns in a children's drawings collection
No full textThe success of large-scale digitization projects at museums, archives, and libraries is pushing other cultural institutions to embrace digitization to preserve their collections. By juxtaposing digital tools with digitized collections, it is now possible to study these cultural objects at a previously unknown scale. This thesis is the first attempt to explore a recently digitized children's drawings collection while developing a system to identify patterns in them linked with popular cultural objects. Artists, as young as three and as old as 25, created nearly 90,000 drawings in the span of three decades from most countries in the world. The preliminary examination unveils that these drawings mirror a solid cultural ethos by using specific iconographic subjects, objects, and colors, and the distinction between children of different parts of the globe is visible in their works. These factors not only make the dataset distinct from other sketch datasets but place it distantly from them in terms of size and multifariousness of creations and the creators. The essential and another dimension of the project is matching the drawings and the popular cultural objects they represent. A deep learning model that learns a metric to rank the visual similarity between the images is used to identify the drawing-artwork pairs. Though the networks developed for image classification perform inadequately for the matching task, networks used for pattern matching in paintings show good performance. Fine-tuning the models increases the performance drastically. The primary outcomes of this work are (1) systems trained with a few methodically chosen examples perform comparably to the systems trained on thousands of generic samples and (2) using drawings enriched by adding generic effects of watercolor, oil painting, pencil sketch, and texturizing mitigates the situation of network learning examples by heart.DHM
Optimal Adaptive Droop Design via a Modified Relaxation of the OPF
No full textThe ever-increasing penetration of renewable energy resources (RESs) in power distribution networks has brought, among others, the challenge of maintaining the grid voltages within the secure region. Employing droop voltage regulators on the RES's inverters is an efficient and low-cast solution to reach this objective. However, fixing droop parameters or optimizing them only for overvoltage conditions does not provide the required robustness and optimality under changing operating conditions. In this article, a convex optimization approach is proposed for reconfiguring P-V and Q-V droop regulators during online operation. The objective is to minimize power curtailment, power losses, and voltage deviation subject to electrical security constraints. This enables to optimally operate the grid with high RES penetration under variable conditions while preserving electrical security constraints (e.g., current and voltage limits). As a first contribution, a mixed-integer linear model of the droop characteristics is developed. According to this model, a droop-regulated generation unit is represented as a constant-power generator in parallel with a constant-impedance load. As a second contribution, a modified augmented relaxed optimal power flow (MAR-OPF) formulation is proposed to guarantee that the electrical security constraints are respected in the presence of constant-impedance loads in the network. Sufficient conditions for the feasibility of the MAR-OPF solution are provided. Those conditions can be checked a priori and are valid for several real distribution networks. Furthermore, an iterative approach is proposed to derive an approximate solution to the MAR-OPF that is close to the global optimal one. The performance of the MAR-OPF approach and the accuracy of the proposed model are evaluated on standard 34- and 85-bus test networks.SCI-STI-GF
Perturbative construction of mean-field equations in extensive-rank matrix factorization and denoising
No full textFactorization of matrices where the rank of the two factors diverges linearly with their sizes has many applications in diverse areas such as unsupervised representation learning, dictionary learning or sparse coding. We consider a setting where the two factors are generated from known component-wise independent prior distributions, and the statistician observes a (possibly noisy) component-wise function of their matrix product. In the limit where the dimensions of the matrices tend to infinity, but their ratios remain fixed, we expect to be able to derive closed form expressions for the optimal mean squared error on the estimation of the two factors. However, this remains a very involved mathematical and algorithmic problem. A related, but simpler, problem is extensive-rank matrix denoising, where one aims to reconstruct a matrix with extensive but usually small rank from noisy measurements. In this paper, we approach both these problems using high-temperature expansions at fixed order parameters. This allows to clarify how previous attempts at solving these problems failed at finding an asymptotically exact solution. We provide a systematic way to derive the corrections to these existing approximations, taking into account the structure of correlations particular to the problem. Finally, we illustrate our approach in detail on the case of extensive-rank matrix denoising. We compare our results with known optimal rotationally-invariant estimators, and show how exact asymptotic calculations of the minimal error can be performed using extensive-rank matrix integrals.IDEPHICS1IDEPHICS2SPOC1SPOC
SPAHM: the spectrum of approximated Hamiltonian matrices representations
No full textPhysics-inspired molecular representations are the cornerstone of similarity-based learning applied to solve chemical problems. Despite their conceptual and mathematical diversity, this class of descriptors shares a common underlying philosophy: they all rely on the molecular information that determines the form of the electronic Schrödinger equation. Existing representations take the most varied forms, from non-linear functions of atom types and positions to atom densities and potential, up to complex quantum chemical objects directly injected into the ML architecture. In this work, we present the spectrum of approximated Hamiltonian matrices (SPAHM) as an alternative pathway to construct quantum machine learning representations through leveraging the foundation of the electronic Schrödinger equation itself: the electronic Hamiltonian. As the Hamiltonian encodes all quantum chemical information at once, SPAHM representations not only distinguish different molecules and conformations, but also different spin, charge, and electronic states. As a proof of concept, we focus here on efficient SPAHM representations built from the eigenvalues of a hierarchy of well-established and readily-evaluated “guess” Hamiltonians. These SPAHM representations are particularly compact and efficient for kernel evaluation and their complexity is independent of the number of different atom types in the database.LCM
Les "Grandes Halles Voyageurs", une architecture durable
No full textÀ la fois abris fonctionnels et instruments de prestige, confortables et esthétiques, les grandes halles couvrant les quais au sein des gares ap- paraissent en même temps que les premiers bâtiments destinés à l’at- tente des voyageurs. Ce patrimoine est devenu un emblème de l’industrialisation du territoire et de l’architecture en fer au XIXe siècle, lieu d’innovations et de perfectionne- ments dans les techniques de construction intervenant en parallèle des Expositions universelles qui se succèdent. Malgré la perte d’une partie de l’ensemble architectural historique au cours du XXe siècle, les « grandes halles voyageurs » restent un élément familier du réseau français. Cet équipement généralement réservé aux gares d’une certaine importance est un marqueur identitaire pour les villes se situant hier ou aujourd’hui au centre d’une étoile ferroviaire. Gestionnaire des gares, SNCF Gares & Connexions et sa filiale AREP main- tiennent les halles en état d’exploitation, favorisant la permanence des fonctions et des usages, soulignant la durabilité du patrimoine architectural ferroviaire ainsi que sa capacité d’adaptation. Le partage de leur histoire et de leurs caractéristiques doit permettre de les inscrire encore davantage dans les enjeux sociétaux et environnementaux actuels.LHS
A polygenic risk score to predict sudden cardiac arrest in patients with cardiovascular disease
No full textUPFELLAYMeeting abstrac
Search for the decay B-0 -> phi mu(+) mu(-)
No full textA search for the decay B-0 -> phi mu(+) mu(-) is performed using proton-proton collisions at centre-of-mass energies of 7, 8, and 13 TeV collected by the LHCb experiment and corresponding to an integrated luminosity of 9 fb(-1). No evidence for the B-0 -> phi mu(+) mu(-) decay is found and an upper limit on the branching fraction, excluding the 0 and charmonium regions in the dimuon spectrum, of 4.4 x 10(-3) at a 90% credibility level, relative to that of the B-s(0) -> phi mu(+) mu(-) decay, is established. Using the measured B-s(0) -> phi mu(+) mu(-) branching fraction and assuming a phase-space model, the absolute branching fraction of the decay B-0 -> phi mu(+) mu(-) in the full q(2) range is determined to be less than 3.2 x 10(-9) at a 90% credibility level.LPH
Unified theory of atom-centered representations and message-passing machine-learning schemes
No full textData-driven schemes that associate molecular and crystal structures with their microscopic properties share the need for a concise, effective description of the arrangement of their atomic constituents. Many types of models rely on descriptions of atom-centered environments, which are associated with an atomic property or with an atomic contribution to an extensive macroscopic quantity. Frameworks in this class can be understood in terms of atom-centered density correlations (ACDC), which are used as a basis for a body-ordered, symmetry-adapted expansion of the targets. Several other schemes that gather information on the relationship between neighboring atoms using "message-passing " ideas cannot be directly mapped to correlations centered around a single atom. We generalize the ACDC framework to include multi-centered information, generating representations that provide a complete linear basis to regress symmetric functions of atomic coordinates, and provide a coherent foundation to systematize our understanding of both atom-centered and message-passing and invariant and equivariant machine-learning schemes. Published under an exclusive license by AIP Publishing.COSM