Diplôme honorifique | EPFL Graph Search

vignette| Le doctorat honoris causa reçu par Jimmy Wales de l'Université de Maastricht (2015). Un diplôme honorifique est un diplôme universitaire pour lequel une université (ou un autre établissement décernant un diplôme) a renoncé à toutes les exigences habituelles, telles que l'inscription, la fréquentation, les crédits de cours, une thèse et la réussite d'examens de synthèse. Il est également connu sous les expressions latines honoris causa (« pour l'honneur ») ou ad honorem (« à l'honneur »). Le diplôme est généralement un doctorat ou, plus rarement, une maîtrise, et peut être décerné à une personne qui n'a aucun lien préalable avec l'établissement universitaire ou qui n'a pas fait d'études post-secondaires antérieures. Voici un exemple d'identification d'un récipiendaire de ce prix: Doctorat en administration des affaires (Hon. Causa). Le diplôme est souvent conféré comme un moyen d'honorer les contributions d'un visiteur distingué à un domaine spécifique ou à la société en gé

Robust discrete choice models with t-distributed kernel errors

Prateek Bansal, Michel Bierlaire, Thomas Gasos

Models that are robust to aberrant choice behaviour have received limited attention in discrete choice analysis. In this paper, we analyse two robust alternatives to the multinomial probit (MNP) model. Both alternative models belong to the family of robit models, whose kernel error distributions are heavy-tailed t-distributions. The first model is the multinomial robit (MNR) model in which a generic degrees of freedom parameter controls the heavy-tailedness of the kernel error distribution. The second alternative, the generalised multinomial robit (Gen-MNR) model, has not been studied in the literature before and is more flexible than MNR, as it allows for alternative-specific marginal heavy-tailedness of the kernel error distribution. For both models, we devise scalable and gradient-free Bayes estimators. We compare MNP, MNR and Gen-MNR in a simulation study and a case study on transport mode choice behaviour. We find that both MNR and Gen-MNR deliver significantly better in-sample fit and out-of-sample predictive accuracy than MNP. Gen-MNR outperforms MNR due to its more flexible kernel error distribution. Also, Gen-MNR gives more reasonable elasticity estimates than MNP and MNR, in particular regarding the demand for under-represented alternatives in a class-imbalanced dataset.

EPFL2020

A Study on the Characteristics of Cu-Mn-Dy Alloy Resistive Thin Films

Ho-Yun Lee

Cu-Mn-Dy resistive thin films were prepared on glass and Al2O3 substrates, which was achieved by co-sputtering the Cu-Mn alloy and dysprosium targets. The effects of the addition of dysprosium on the electrical properties and microstructures of annealed Cu-Mn alloy films were investigated. The composition, microstructural and phase evolution of Cu-Mn-Dy films were characterized using field emission scanning electron microscopy, transmission electron microscopy and X-ray diffraction. All Cu-Mn-Dy films showed an amorphous structure when the annealing temperature was set at 300 degrees C. After the annealing temperature was increased to 350 degrees C, the MnO and Cu phases had a significant presence in the Cu-Mn films. However, no MnO phases were observed in Cu-Mn-Dy films at 350 degrees C. Even Cu-Mn-Dy films annealed at 450 degrees C showed no MnO phases. This is because Dy addition can suppress MnO formation. Cu-Mn alloy films with 40% dysprosium addition that were annealed at 300 degrees C exhibited a higher resistivity of 2100 cm with a temperature coefficient of resistance of -85 ppm/degrees C.

2019

Engineering Parallel Transmit/Receive Radio-Frequency Coil Arrays for Human Brain MRI at 7 Tesla

Jérémie Daniel Clément

Magnetic resonance imaging is widely used in medical diagnosis to obtain anatomical details of the human body in a non-invasive way. Clinical MR scanners typically operate at a static magnetic field strength (B0) of 1.5T or 3T. However, going to higher field is of great interest since the signal-to-noise ratio is proportional to B0. Therefore, higher image resolution and better contrast between the human tissues could be achieved. Nevertheless, new challenges arise when increasing B0. The wavelength associated with the radio-frequency field B1+ has smaller dimensions - approx. 12 cm for human brain tissues - than the human brain itself (20 cm in length), the organ of interest in this thesis. The main consequence is that the transmit field distribution pattern (B1+) is altered and the final MR images present bright and dark signal spots. These effects prevent the ultra-high field MR scanners (>= 7T) to be used for routine clinical diagnosis. Parallel-transmit is one approach to address these new challenges. Instead of using an RF coil connected to a single power input as it is commonly done at lower magnetic fields, multiple RF coils are used with independent power inputs. The subsequent distinct RF signals can be manipulated separately, which provides an additional degree of freedom to generate homogeneous B1+-field distributions over large or specific regions in the human body. A transmit/receive RF coil array optimized for whole-brain MR imaging was developed and is described in this thesis. Dipoles antennas were used since they could provide a large longitudinal (vertical axis-head to neck) coverage and high transmit field efficiency. Results demonstrated a complete coverage of the human brain, and particularly high homogeneity over the cerebellum. However, since the receive sensitivity over large field-of-views is related to the number of channels available to detect the NMR signal, the next work was to add a 32-channel receive loop coil array to the transmit coil array. The complete coverage of the human brain was assessed with a substantial increase in signal-to-noise compared to the transmit/receive dipole coil array alone. Moreover, acquisition time was shortened since higher acceleration factors could be used. To optimize the individual RF fields and generate an homogeneous B1+-field, a method was developed making use of the particle-swarm algorithm. A user-friendly graphical interface was implemented. Good homogeneity could be achieved over the whole-brain after optimization with the coil array built in this study. Moreover, the optimization was shown to be robust across multiple subjects. The last project was focused on the single transmit system. Local volume coils (single transmit) present pronounced transmit field inhomogeneities in specific regions of the human brain such as the temporal lobes. A widely used approach to address locally these challenges is to add dielectric pads inside the volume coils to enhance the local transmit field efficiency. It was shown in this thesis that constructing dedicated surface coils is a valuable alternative to the dielectric pads in terms of transmit field efficiency and MR spectroscopy results. Two RF coil setups were developed for the temporal and frontal lobes of the human brain, respectively. This thesis provides extensive insight on MR engineering of RF coils at ultra-high field and the potential of parallel-transmit to address the future needs in clinical applications.

EPFL2019

Engineering Parallel Transmit/Receive Radio-Frequency Coil Arrays for Human Brain MRI at 7 Tesla

Jérémie Daniel Clément

EPFL2019