Mars - Juin - Octobre


Mars : Jeudi 13 mars 2014 de 14h00-16h00

Thème « Simulation conditionnelle et estimation bayesienne » :

1. Lieu :
Mines ParisTech, Bd Saint-Michel
salle  L-218.

2. Descriptif :

PREMIER EXPOSé : 14h00-15h00

Orateur : Christian Lantuéjoul (centre de Géosciences, christian.lantuejoul@mines-paristech.fr)

Titre : Simulation conditionnelle de modèles max-stables

Résumé : Les champs aléatoires max-stables permettent de modéliser des phénomènes extrêmes dans de nombreux domaines de l'environnement ou du climat. N'ayant pas de moments finis aux deux premiers ordres, les techniques usuelles de krigeage ne peuvent pas être utilisées pour estimer des risques ou envisager les conséquences de divers scénarios. Dans une telle situation, il est utile de recourir à des simulations conditionnelles. Partant d'un algorithme de simulation conditionnelle mis au point par Dombry, Eyi-Minko et Ribatet (Biometrika, Vol. 20, pp. 1-23, 2012), un nouvel algorithme est proposé qui s'appuie sur des variables latentes et qui s'avère plus rapide, plus précis et plus apte à traiter de grands jeux de données. Ce travail, financé par l'ANR Mc-SIM, s'effectue en colloboration avec L. Bel (AgroParisTech) et J.N. Bacro (Univ. Montpellier 2).

SECOND EXPOSé : 15h00-16h00

Speaker : Thomas Romary (centre de Géosciences, thomas.romary@mines-paristech.fr)

Title : Bayesian seismic tomography by parallel interacting Markov chains

Abstract : First arrival travel time tomography aims at estimating the velocity field of the subsurface. The resulting velocity field is then commonly used as a starting point for further seismological, mineralogical or tectonic analysis in a wide range of applications such as geothermal energy, volcanoes studies,... The estimated velocity field is obtained through inverse modeling by minimizing an objective function that compares observed and modeled travel times. This step is usually performed by steepest descent optimization algorithms. The major flaw of such optimization schemes, beyond the possibility of staying stuck in a local minima, is that they do not account for the multiple possible solutions of the inverse problem at stake. Therefore, they are unable to assess the uncertainties linked to the solution. In a Bayesian perspective however, the inverse problem can be seen as a conditional simulation problem, that can be solved using iterative algorithms. Indeed, Markov chains Monte-Carlo (MCMC) methods are known to produce samples of virtually any distribution. They have already been widely used in the resolution of non-linear inverse problems where no analytical expression for the forward relation between data and model parameters is available, and where linearization is unsuccessful. In Bayesian inversion, the total number of simulations we can afford is highly related to the computational cost of the forward model. Although fast algorithms have been recently developed for computing first arrival travel time of seismic waves, the complete browsing of the posterior distribution is hardly performed at final time, especially when it is high dimensional and/or multimodal. In the latter case, the chain may stay stuck in one of the modes. One way to improve the mixing properties of classical single MCMC is by making interact several Markov chains at different temperatures. These methods can make efficient use of large CPU clusters, without increasing the global computational cost with respect to classical MCMC and are therefore particularly suited for Bayesian inversion. The approach is illustrated on two geophysical case studies.


Juin : Jeudi 12 juin 2014 de 14h00-16h00

Thème « Feedback à l'échelle quantique » :

1. Lieu :
Mines ParisTech, Bd Saint-Michel
salle  V106-A.

2. Descriptif :

PREMIER EXPOSÉ : 14h00-15h00

Speaker :François Mallet (laboratoire Pierre Aigrain, ENS Paris, francois.mallet@lpa.ens.fr)

Title : Persistent control of a superconducting qubit by stroboscopic measurement feedback

Abstract : Recently, research on quantum physics has moved from only demonstration of subtle quantum phenomena to engineered systems benefiting from quantum laws peculiarities. Quantum measurements being non-deterministic, it is mandatory to develop feedback procedures in the aim of building predictable quantum machines. Feedback is made possible by the fact that the measurement record contains all the necessary information to follow the quantum state of the system in real time. In this talk, I will present a recent experiment where a superconducting qubit is controlled by a projective measurement based feedback. We demonstrate efficient qubit reset and trajectory stabilization using the information extracted from a Quantum Non Demolition projective measurement. This experiment benefits from 3 recent breakthroughs. First, we use the recently developed 3D transmon superconducting qubits, with several tens of microseconds coherence times [1].
Second, the detector for the qubit pointer states, which are a few microwave photons large coherent states, is a phase preserving parametric amplifier operating close to the quantum limit and recently developed in our group [2].
Third, the feedback loop can be performed with a delay of a fraction of a microsecond using a digital filter implemented on innovative fast electronic digital processors, namely Field-Programmable-Gate-Arrays (FPGA). In this experiment we combine these three ingredients to realize a test bed for measurement based feedback.
As a first illustration of its abilities, we demonstrate high fidelity qubit state preparation, approx. 99% for the ground state and approx. 95% for the excited state. We also demonstrate the perpetual stabilization of Rabi oscillations, with fidelity above 80% [3].

[1] H. Paik et al., Phys. Rev. Lett. 107, 240501 (2011)
[2] N. Roch et al., Phys. Rev. Lett.108, 147701 (2012)
[3] P. Campagne-Ibarcq et al., Phys. Rev. X. 3, 021008 (2013)

SECOND EXPOSÉ : 15h00-16h00

Orateur : Pierre Rouchon (centre automatique et systèmes , pierre.rouchon@mines-paristech.fr)

Titre : Modèles et stabilisation par feedback de systèmes quantiques ouverts

Résumé : Au niveau quantique, le feedback doit prendre en compte l'action en retour attachée à la mesure d'une observable. Nous présentons ici la structure des modèles markoviens incluant cette action en retour ainsi que deux méthodes de stabilisation par feedback: le feedback fondé sur la mesure où un système quantique est stabilisé par un contrôleur classique; le feedback cohérent, dit aussi autonome et proche de l'ingénierie de réservoir, où un système quantique est stabilisé par un contrôleur lui aussi quantique. Ces modèles et méthodes sont d'abord expliqués pour la boite à photons du LKB. Ils sont ensuite présentés pour un système quantique ouvert arbitraire.


Octobre : Jeudi 09 octobre 2014 de 14h00-16h00

Thème « Mobile mapping » :

1. Lieu :
Mines ParisTech, Bd Saint-Michel
salle  L118.

2. Descriptif :

PREMIER EXPOSÉ : 14h00-15h00

Speaker : Daniela Craciun (CAOR, daniela.craciun@mines-paristech.fr)

Title : Scalable and detail-preserving ground surface reconstruction from large 3d point clouds acquired by mobile mapping systems

Abstract: The currently existing mobile mapping systems equipped with active 3D sensors allow to acquire the environment with high sampling rates at high vehicle velocities. While providing an effective solution for environment sensing over large scale distances, such acquisition provides only a discrete representation of the geometry. Thus, a continuous map of the underlying surface must be built.
Mobile acquisition introduces several constraints for the state-of-the-art surface reconstruction algorithms.
Smoothing becomes a difficult task for recovering sharp depth features while avoiding mesh shrinkage. In addition, interpolation based techniques are not suitable for noisy datasets acquired by Mobile Laser Scanning (MLS) systems. Furthermore, scalability is a major concern for enabling real-time rendering over large scale distances while preserving geometric details.

This paper presents a fully automatic ground surface reconstruction framework capable to deal with the aforementioned constraints. The proposed method exploits the quasi-flat geometry of the ground throughout a morphological segmentation algorithm. Then, a planar Delaunay triangulation is applied in order to reconstruct the ground surface. A smoothing procedure eliminates high frequency peaks, while preserving geometric details in order to provide a regular ground surface. Finally, a decimation step is applied in order to cope with scalability constraints over large scale distances. Experimental results on real data acquired in large urban environments are presented and a performance evaluation with respect to ground truth measurements demonstrate the effectiveness of our method.

References: D. Craciun, A. M. Serna, J-E. Deschaud, B. Marcotegui and F. Goulette. Scalable and detail-preserving ground surface reconstruction from large 3D point clouds acquired by Mobile Mapping Systems. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XL-3, 2014. ISPRS Technical Commission III Symposium, 5 – 7 September 2014, Zurich, Switzerland.

SECOND EXPOSÉ : 15h00-16h00

Speaker : Andrés Serna (CMM: andres.serna_morales@mines-paristech.fr)

Title : Semantic analysis of 3D point clouds from urban environments using mathematical morphology

Abstract : In the last ten years, thanks to new 3D data availability, an increasing number of geographic applications such as Google Earth, Geoportail, iTowns and Elyx-3D is flourishing. Most of them are enhanced with pedestrian navigation options and realistic 3D models. In order to build realistic and faithful to reality 3D maps, data need to be processed in a large scale way. This processing usually consists in transforming points into surfaces or geometric primitives for subsequent analysis and modeling.
This stage is often manual and therefore time consuming. Thus, object extraction from urban scenes is difficult and tedious, and existing semi-automatic methods may not be sufficiently precise nor robust. An automatic semantic analysis of extracted objects combined with the intelligent intervention of an operator would speed up and improve the process.

My Ph.D. thesis entitled "Semantic analysis of 3D point clouds from urban environments using mathematical morphology" is developed in the framework of TerraMobilita project. The general goal is the development of new automated methods for 3D urban cartography. In particular, we aim at developing automatic methods to process 3D point clouds from urban laser scanning. Our methods are based on elevation images, mathematical morphology and supervised learning. Although the processing of 3D urban data has been underway for many years, automatic semantic analysis is still an open research problem. The development of accurate and fast algorithms in this domain is one of the main contributions of this thesis.

In this seminar, two of the thesis contributions are presented:

1- Semantic analysis of 3D urban objects: in order to detect, segment and classify urban objects. In this process, an identifier and a class are assigned to each segmented object. Each class represents an urban semantic entity. Depending on the application, several classes can be defined, e.g. lampposts, traffic signs, benches, cars, pedestrians, garbage containers, bikes, among others. This separation is be useful to produce detailed 3D urban maps, to define the best itinerary for a specific mobility type, to produce parking areas maps and to compute parking statistics.

2- Ground segmentation and accessibility analysis: in order to define a Digital Terrain Model (DTM) with extra features such as access ramps geometry, useful to establish the suitability of a path for a specific mobility type. For example, high curbs should be avoided for a skater or a person in a wheelchair.

Our methods have been favorably compared with state of the art methods on public 3D databases.