The Chesapeake Bay impact structure is one of the largest and most well preserved impact structures on Earth. It has a unique morphology composed of an inner crater penetrating crystalline basement surrounded by a wider crater in the overlying sediments. In 2004, the U.S. Geological Survey conducted a seismic survey with the goals of constraining crater structure and in support of the drilling of a borehole into the deepest part of the crater. Travel-time and waveform inversion were applied to the data to produce a high-resolution velocity model of the crater. Low-fold reflection processing was also applied. Northeast of the crystalline crater, undeformed, eastward-sloping crystalline basement is ~1.5 km deep. The edge of the inner crater is at ~ 15 km radius and slopes gradually down to a depth of 1.5 - 1.8 km. A central peak of 4-5 km radius rises to a depth of ~0.8 km. Basement velocity in the crystalline crater is much lower than undeformed basement, which suggests ~10% fracturing of the crater floor, and up to 20% fracturing of the central uplift. A basement uplift and lateral change of velocity, interpreted as the edge of the transient crater, occurs at a radius of ~ 11 km. Assuming a 22 km diameter transient crater, scaling laws predict a ~30 km diameter crater and central peak diameter of 8-10 km. This indicates that post-impact collapse processes that created the ~ 30 km diameter crystalline crater were unaffected by the much weaker rheology of the overlying sediments. / Master of Science

Federated learning (FL) is a communication protocol that allows multiple distrustful clients to collaboratively train a machine learning model. In FL, data never leaves client devices; instead, clients only share locally computed gradients or model parameters with a central server. As individual gradients may leak information about a given client's dataset, secure aggregation was proposed. With secure aggregation, the server only receives the aggregate gradient update from the set of all sampled clients without being able to access any individual gradient. One challenge in FL is the systems-level heterogeneity that is quite often present among client devices. Specifically, clients in the FL protocol may have varying levels of compute power, on-device memory, and communication bandwidth. These limitations are addressed by model-heterogeneous FL schemes, where clients are able to train on subsets of the global model. Despite the benefits of model-heterogeneous schemes in addressing systems-level challenges, the implications of these schemes on client privacy have not been thoroughly investigated. In this thesis, we investigate whether the nature of model distribution and the computational heterogeneity among client devices in model-heterogeneous FL schemes may result in the server being able to recover sensitive information from target clients. To this end, we propose two novel attacks in the model-heterogeneous setting, even with secure aggregation in place. We call these attacks the Convergence Rate Attack and the Rolling Model Attack. The Convergence Rate Attack targets schemes where clients train on the same subset of the global model, while the Rolling Model Attack targets schemes where model-parameters are dynamically updated each round. We show that a malicious adversary is able to compromise the model and data confidentiality of a target group of clients. We evaluate our attacks on the MNIST dataset and show that using our techniques, an adversary can reconstruct data samples with high fidelity. / Master of Science / Federated learning (FL) is a communication protocol that allows multiple distrustful users to collaboratively train a machine learning model together. In FL, data never leaves user devices; instead, users only share locally computed gradients or model parameters (e.g. weight and bias values) with an aggregation server. As individual gradients may leak information about a given user's dataset, secure aggregation was proposed. Secure aggregation is a protocol that users and the server run together, where the server only receives the aggregate gradient update from the set of all sampled users instead of each individual user update. In FL, users often have varying levels of compute power, on-device memory, and communication bandwidth. These differences between users are collectively referred to as systems-level (or system) heterogeneity. While there are a number of techniques to address system heterogeneity, one popular approach is to have users train on different subsets of the global model. This approach is known as model-heterogeneous FL. Despite the benefits of model-heterogeneous FL schemes in addressing systems-level challenges, the implications of these schemes on user privacy have not been thoroughly investigated. In this thesis, we investigate whether the nature of model distribution and the differences in compute power between user devices in model-heterogeneous FL schemes may result in the server being able to recover sensitive information. To this end, we propose two novel attacks in the model-heterogeneous setting with secure aggregation in place. We call these attacks the Convergence Rate Attack and the Rolling Model Attack. The Convergence Rate Attack targets schemes where users train on the same subset of the global model, while the Rolling Model Attack targets schemes where model-parameters may change each round. We first show that a malicious server is able to obtain individual user updates, despite secure aggregation being in place. Then, we demonstrate how an adversary can utilize those updates to reverse engineer data samples from users. We evaluate our attacks on the MNIST dataset, a commonly used dataset of handwritten digits and their labels. We show that by running our attacks, an adversary can accurately identify what images a user trained on.

The propagation of radiowaves through ionospheric irregularities can lead to random amplitude and phase fluctuations of the signal, otherwise known as scintillation, which can severely impact the performance of Global Navigation Satellite System (GNSS) and communication systems. Research into high latitude scintillation, through statistical analysis and inverse modeling, was completed to provide insight into the temporal and spatial distribution, and irregularity parameters, which can ultimately support the development of impact mitigation techniques, and deepen our understanding of the underlying physics. The work in this dissertation focused on the statistical analysis of Global Positioning System (GPS) scintillation data, data inversion, two-dimensional (2D) and three-dimensional (3D) scintillation modeling. The statistical analysis revealed distinct trends in the distribution of scintillation, while demonstrating that for GPS signals, phase scintillation occurs most frequently and can be treated as stochastic Total Electron Content (TEC); findings which have significant implications for impact mitigation. For the first of two inversion studies, scintillation data associated with a series of Polar Cap Patches (PCPs), which are common large-scale high latitude structures, was inverted to gain insight into the composition of the underlying irregularities. The results of this study suggest that the irregularities can be modeled as rods interbedded with sheets, which is knowledge that is crucial for the anchoring of models used to develop system mitigation techniques. The final study presents the results of modeling and inversion work to identify the conditions under which a 2D analytic version of the 3D numerical Satellite-beacon Ionospheric-scintillation global model of the upper atmosphere (SIGMA) model can be used to perform modeling in high latitude regions. During the study, it was found that the analytic model tends to diverge for electron density variance times irregularity layer thickness values exceeding 2, matched reasonably well for correlation length to thickness ratios up to 0.2, and was incompatible when ratios approached 0.35. An elevation angle limitation was also identified for the 2D model, and inflated values for the electron density variance were observed overall, which are thought to result from the weak scatter limits of the analytic model. These inflated values were particularly acute in the auroral zone during elevated conditions and suggest that the analytic model used in the study is not well suited for modeling the highly elongated irregularities associated with auroral precipitation. / Doctor of Philosophy / The ionosphere is a region of the earth's atmosphere extending from approximately 90 to 1000 km in altitude. Radio wave signals which travel through irregularities in the ionosphere can be distorted in a way that can lead to random amplitude and phase fluctuations of the signal, otherwise known as scintillation, which can severely degrade the performance of navigation and communication systems. Research into high latitude scintillation, through statistical analysis, and data and model matching, was completed to provide insight into the time and space distribution, and irregularity parameters, in order to ultimately deepen our understanding of the physics and to help develop better models. The work in this dissertation focused on the statistical analysis of GPS scintillation data, data and model matching, and 2D and 3D irregularity modeling. The statistical analysis revealed distinct trends in the distribution of scintillation, while demonstrating that for GPS signals, phase scintillation occurs most frequently but the impacts can be corrected if measured; findings which have significant implications for impact mitigation. For the first of two model and data matching studies, scintillation data associated with a series of common large-scale high latitude structures called PCPs, was matched to a model to gain insight into the composition of the underlying irregularities. The results of this study suggest that the irregularities can be modeled as vertical rods oriented along the magnetic field interbedded within flat sheets, which is knowledge that is crucial for having confidence in the models used to develop system mitigation techniques. The final study presents the results of modeling and data matching work to identify the conditions under which a 2D or 3D model can be used to perform irregularity modeling in the high latitude regions. During the study, it was found that the 2D model tends to diverge from the 3D model for significant variations in the ionosphere, and when irregularity rods are highly elongated. A signal propagation path elevation angle limitation was also identified for the 2D model, and inflated values for the predicted ionospheric variations were observed overall, which are thought to result from limits of the 2D model compared to the more general 3D version. These inflated values were particularly acute in the auroral region during elevated conditions and suggest that the 2D model used in the study is not well suited for modeling the highly elongated irregularities associated with aurora effects.

Detta examensarbete fokuserar på associativitet inom matematikdidaktisk forskning. Fram till sent 90-tal har associativitet fått begränsad uppmärksamhet, vilket resulterar i relativt lite forskning jämfört med andra matematiska egenskaper. Trots att associativitet ligger till grund för flera beräkningsstrategier som undervisas i grundskolan, är förståelsen för denna egenskap ofta låg bland eleverna. Därför är intresset med vårt arbete att utforska metoder för att öka elevernas förståelse för associativitet och därigenom förbättra deras kunskaper inom matematik.  Syftet med studien är att belysa med vilka ord som används för att beskriva associativitet i matematikdidaktisk forskning och dess relation med inversion. Genom systematiska sökningar i databaser identifierades 14 vetenskapliga artiklar för analys. Artiklarna analyserades med en komparativ metod för att identifiera och synliggöra hur associativitet behandlas inom dessa artiklar. Resultatet visar att associativitet beskrivs på olika sätt: som en genväg, strategi, genvägsstrategi och konceptuellt baserad strategi. Dessutom betraktas det som en sällan använd problemlösningsmetod bland elever. Det visar även att det finns en stark relation mellan inversion och associativitet. Studien framhäver också en stark koppling mellan inversion och associativitet. Slutsatsen är att artiklarna inte enbart diskuterar associativitet utan snarare en metod som bygger på och underlättar förståelsen av associativitet. Metoden visar sig vara mer effektiv än den traditionella vänster-till högerstrategin för att göra beräkningar i form utav  dvs. Uttryck med tre termer där den andra operationen är räknesättens inverser.

As the dimension and the complexity of large interconnected systems grow, so does the necessity for decentralized control. One of the interesting challenges in the field of decentralized control is the arbitrary pole placement using output feedback. The feasibility of this problem depends solely on the identification of the decentralized fixed modes. As a matter of fact, if the system is free of fixed modes, then by increasing the controller’s order, any arbitrary closed loop poles can always be assigned. Due to this fact, reducing the controller’s order constitutes another interesting challenge when dealing with decentralization. This research describes the decentralized pole placement of linear systems. It is assumed that the internal structure of the system is unknown. The only access to the system is from a number of control stations. The decentralized controller consists of output feedback controllers each built at a control station. The research can be divided into two parts. In the first part, conditions for fixed modes existence as well as realization and stability of the overall system under decentralization are established using polynomial matrix algebra. The second part deals with the solution of decentralized pole placement problem, in particular, finding a decentralized controller which assigns some set of desired poles. The solution strategy is to reduce the controller’s order as much as possible using mathematical programming techniques. The idea behind this method is to start with a low order controller and then attempt to shift the poles of the closed loop system to the desired poles. / Ph. D. / incomplete_metadata

Résumé : Le vieillissement des ouvrages en béton est une préoccupation majeure affectant la pérennité et l’efficacité des structures. Le maître d’ouvrage se doit de maintenir les fonctions d’usage de la structure tout en gardant une gestion économique efficace. L’objectif final de ces travaux de recherche est, donc de pouvoir renseigner sur l’état global de fissuration de la structure afin d’aider le maître d’ouvrage à respecter ses engagements. Dans cette optique, cette thèse développe une nouvelle technique aidant à la quantification de l’état des ouvrages massifs en béton. Elle s’appuie, pour cela, sur la méthode non-destructive de résistivité électrique en surface, connue pour sa sensibilité face à des facteurs révélateurs d’une altération. Toutefois, à cause de sa dépendance entre la profondeur d’investigation et la résolution, la méthode ne peut pas garantir de l’état global d’un ouvrage. De ce fait, il a été décidé d’utiliser la résistivité électrique via des forages préexistants dans la structure (diagraphie électrique). L’outil utilisé est une sonde en dispositif normal réservée jusqu’à présent pour la prospection pétrolière et hydrogéologique. En plus d’une prospection en profondeur via le forage, cette sonde peut acquérir des informations sur un rayon de 3.2m autour du forage. Cependant, à mesure que le volume de béton sondé augmente, la résolution décroit. La difficulté est donc de pouvoir exploiter les capacités de prospection de la sonde tout en sachant que la résolution faillit. Il s’agit de contourner le problème en maîtrisant les concepts de la diagraphie et son nouveau milieu d’application. Cette thèse est basée sur une première approche numérique permettant d’apporter des corrections sur les données de terrain et de déterminer la sensibilité de l’outil face à de l’endommagement d’ouverture plurimillimétrique à centimétrique. Ceci est validé par des mesures réalisées sur une écluse de la Voie Maritime du Saint-Laurent. Une étude numérique de la réponse de l’outil en fonction des paramètres de fissure tels que l’ouverture, le contraste entre la résistivité de la discontinuité et du béton, et l’extension est réalisée. Elle permet de construire une base de données afin de développer une méthode pour la caractérisation de l’endommagement. Cette méthode s’appuie sur ces réponses diagraphiques pour retrouver les paramètres de fissure recherchés (problème inverse). Nous procédons tout d’abord par une analyse préliminaire se basant sur un croisement des informations apportées par les différentes électrodes de la sonde puis nous optimisons les résultats par la méthode de recuit simulé. La méthode, ainsi développée est ensuite appliquée à un deuxième ouvrage pour en déterminer l’état interne. Ces travaux détectent plusieurs zones endommagées et caractérisent l’une d’elles par une ouverture centimétrique et une extension comprise entre 1.6m et 3.2m. Ces travaux prometteurs, attestent d’un premier diagnostic interne des ouvrages massifs en béton, un enjeu qui restait sans réponses satisfaisantes jusqu’à maintenant. // Abstract : The aging of concrete structures is a major problem affecting their sustainability and their efficiency. The owner must maintain the structure serviceability and provide cost-effective management. The goal of this work is to provide detailed information about the state of cracking inside the structure in order to assist the owner to meet its commitments. In this context, this thesis develops a new technology to assess the condition of mass concrete structures. It relies on a non-destructive method based on electrical resistivity measured from surface, known for its sensitivity to factors associated with concrete deterioration. However, because of its dependence between the investigation depth and the resolution, the method cannot assess the overall state of a structure. Therefore, it was decided to use the electrical resistivity through preexisting boreholes in the structure (electrical logging). The tool used is a normal probe, which has been traditionally used for oil and hydrogeological exploration. In addition to the investigation in depth via boreholes, this probe can get information over a radius of 3.2m around the borehole. However, as the probing volume of concrete increases, the resolution decreases. Difficulty is to use the exploration abilities of the tool, knowing that the resolution is limited. This is to get around the problem by mastering logging concepts and its new application environment. This thesis is based on a first numerical approach to make corrections on field data and to determine the tool sensitivity with regard to the multi-millimeter and centimeter crack size damage. This was validated with measurements made on a full-size lock located on the St. Lawrence Seaway. A numerical study of the tool response versus the discontinuities parameters such as the crack aperture, the resistivity contrast between the discontinuity and the concrete, and the extension was done. It allowed building a database used to develop a method for the characterization of the damage. This method is based on the tool responses to find the crack parameters (inverse problem). First, we proceed with a preliminary analysis based on a cross of information provided by the different electrodes of the probe then we optimize the results by the method of simulated annealing. The characterization method is applied to another structure to quantify its internal state. These studies detect several damaged areas and characterize one of them by a centimeter aperture and an extension between 1.6m and 3.2m. This work attest to a first internal diagnosis of massive concrete structures, an issue that remained without satisfactory answers so far.

In this thesis, four studies using different geophysical electromagnetic methods are presented. In the first study dealing with airborne measurements, the noise response due to the rotation of the aircraft and the aircraft itself as a metallic conductive body on the Earth's electromagnetic response in very low frequency and low frequency band was investigated. The magnetic fields are independent of the aircraft in the VLF band and part of the LF band. But at higher frequencies (above 100 kHz), the signals are more influenced by the aircraft. The aircraft also generates its own noise frequencies which are mixed with the radio transmitter signals. The second and third studies are applications of radio-, controlled source-magnetotellurics and electrical resistivity tomography methods at a quick-clay landslide site in southwest Sweden. The data are processed and modeled in 2D and 3D, and the models are compared with high-resolution seismic and geotechnical data. The obtained results were further validated and refined by performing synthetic tests in the second study. The third study shows that the 3D models provide larger and more continuous volume of the quick clay structure than traditional 2D models. Both studies have shown that integrated application of geophysical methods for landslides is ideal. Quick clays often overlie the coarse-grained layers showing an increase of resistivity values in the models. In the fourth study, a new audio magnetotelluric data acquisition technique is developed and is named moving magnetotellurics (MMT). In this new technique, the magnetic sensors are placed on the ground and only 15 to 20 minutes data are acquired for each station, which usually is enough to cover the frequency range 30-300 Hz. The new technique is more efficient and convenient than the traditional magnetotelluric method, and test measurements have shown that it is an applicable method in shallow depth studies.

Grâce aux progrès expérimentaux permettant d'étudier deux phénotypes qui sont d'un intérêt majeur dans la compréhension des capacités d'adaptation d'A. gambiae s.s. à son environnement. Les différences de résistance à la dessiccation mises en évidence entre les différents caryotypes liés à l'inversion chromosomique 2La et entre les formes moléculaires M et S offrent des pistes intéressantes pour l'identification de facteurs génétiques impliqués dans la divergence écologique au sein de ce complexe d'espèces. / Thanks to progress in sequencing, the genomes of many organisms are known and available. Thus, functional genomics, the elucidation of gene function in sequenced genome, is currently booming. However, there is a gap between our growing knowledge in genetic and the current sparse information on phentoypic data ( "phenotype gap"). All organisms whose genome has been sequenced are facing this problem, including Anopheles gambiae.Anopheles gambiae sensu lato is a complex of sibling species, indistinguishable from a morphological point of view, present on almost the entire African continent. A. gambiae demonstrates an extreme environmental ubiquity and the characterization of phenotypes associated with adaptation to varying environments as well as the identification of genes involved in this adaptation is one of the main research axes in the post-genome area of this major malaria vector.We have studied some phenotypes associated with aridity and temperature in the nominal species of the A. gambiae complex. These two parameters are discriminent in the distribution of molecular forms and chromosomal inversions that characterize this species and may be involved in ecological divergence and speciation. We first measured desiccation resistance of adult mosquitoes of A. gambiae s.s. and we then studied the preferred temperatures of larvae in a choice device set-up (the shuttlebox). We compared the thermoregulation behavior and thermal preferences of a laboratory strain with field larvae of A. gambiae s.s. We also presented preliminary data on the preferred temperatures measured in field larvae of the S and M molecular forms.From a technical point of view, we improved an existing device for testing the survival of mosquitoes in highly desiccated conditions by coupling it with a video surveillance system, which help to increase the accuracy in determining the survival time, to avoid disturbing the system during the experiment and allow to test relatively large numbers of individuals. This study revealed a significant association between the 2La chromosomal inversion and resistance to desiccation in A. gambiae and highlighted the role of body size in the survival of this mosquito in dry environments.We also adapted a new device to study experimentally the thermopreference of A. gambiae s.s. larvae. The results showed that laboratory larvae and field M molecular form larvae had similar thermal preferences, consistent with the values of temperature usually found in natural breeding sites. In addition, the S molecular form larvae from southern Cameroon had preferences similar to those of northern Cameroon, regardless of karyotypes related to chromosomal inversions. In addition, the comparison of data for the M and S molecular forms larvae revealed that there was no significant difference in thermal preferences or in thermoregulatory behavior.Our results have contributed to the development of two experimental devices to study two phenotypes that are of major interest in understanding the adaptation of A. gambiae s.s. to its environment. The differences in desiccation resistance between the different karyotypes associated with the 2La chromosomal inversion and between the M and S molecular forms offer interesting new possibilities for the identification of genetic factors involved in their ecological divergence.

En Méditerranée orientale, la paleo-marge sud de la Téthys a subit des épisodes polyphasés d'extension pendant le Paléozoïque et le Mésozoïque. Cette marge a été postérieurement inversée pendant des épisodes compressifs et discontinus depuis le Crétacé supérieur liés à la convergence entre l'Afrique et l'Eurasie.La marge Cyrénaïque (nord-est Libye) a enregistré ces épisodes extensifs et compressifs. Elle permet donc l'analyse des inversions et de leurs relations avec les évènements ayant eu lieu le long de la frontière de la plaque Africaine (i.e. subduction Hellenique).Le bassin de Sirte, adjacent à la Cyrénaïque montre une direction oblique, ne présente pas la même déformation et est caractérisé par une subsidence continue depuis le Mésozoïque.Des données de sismique (réflexion) combinées à des rapports et des corrélations de puits, nous ont permis d'examiner et de discuter les interactions entre la Cyrénaïque, le bassin de Sirte et les domaines profonds (i.e. bassin Ionien).Ce travail permet de mieux préciser les différents épisodes d'extension, de contraindre les évènements compressifs enregistrés par la région Cyrénaïque, d'observer les structures du bassin de Sirte et de clarifier en partie son évolution.Finalement cette partie de la marge est intégrée dans le cadre géodynamique régional de la Téthys sud et permet de discuter l'âge ainsi que le mécanisme d'ouverture pour la branche orientale de la Méditerranée. / In the Eastern Mediterranean, the South-Tethys paleo-margin experienced poly-phased rifting episodes during Paleozoic and Mesozoic times. This margin has been subsequently inverted by discontinuous events occurring since the Late Cretaceous as a consequence of the Africa-Eurasia convergence.The Cyrenaica margin (northeast Libya) has recorded these extensional and compressional events. It thus gives the opportunity to analyse these inversion and their possible causal links with events occurring along the plate boundary (i.e. within the Hellenic subduction).The adjacent Sirt Basin, follows an oblique direction, did not suffer the same deformation as Cyrenaica and has recorded a continuous subsidence since the Mesozoic.Offshore seismic data combined with well correlations have permitted us to investigate and discuss the interactions between Cyrenaica, Sirt Basin and the deeper domains (i.e. Ionian Basin).We were able to document the different rift episodes, better constrain the compressional events on Cyrenaica, observe characteristics of the architecture of the Sirt Basin and clarify part of its evolution.Finally we integrate this part of the margin, in the regional geodynamic frame of the East Mediterranean branch of the Neo-Tethys by discussing the timing and mechanism which led to its opening.

L’amélioration des images sismiques peut aider à mieux contraindre l’exploration deshydrocarbures. Les ondes élastiques qui se propagent dans la Terre peuvent être classifiéescomme ondes de volume et ondes de surface. Si ces dernières sont les plus énergétiques,seules les ondes de volume sont couramment considérées comme des signaux utiles.Cependant, les ondes de surface sont utiles pour caractériser la proche sub-surface.Classiquement, les ondes de surface sont analysées dans des contextes de propriétésélastiques localement 1D.Nous proposons une modification de l’inversion des formes d’onde classique pourreconstruire des profils de propriétés 2D (la windowed-Amplitude Waveform Inversion, w-AWI). La w-AWI est spécialement robuste en ce qui concerne le choix du modèle initial.Nous appliquons la w-AWI aux données synthétiques ainsi qu’aux données réelles, montrantque cette approche permet de récupérer des propriétés 2D. / High-resolution seismic imaging is essential to improve results of hydrocarbon exploration.Elastic waves propagate in the Earth as body and surface waves, the latter being the mostenergetic ones. Body waves are preferred for exploration seismic imaging while surfacewaves are usually considered to be noise. However, it has been recognised that the nearsurface can be characterised by analysing surface waves and that such result may improvethe outcome of body-wave processing. Currently, surface waves analysis leads to retrievelocal 1D property profiles.We propose a waveform-based inversion procedure to derive 2D velocity models fromsurface waves. This method consists of a misfit functional modification of classical FullWaveform Inversion and we call it windowed-Amplitude Waveform Inversion (w-AWI). Weshow that w-AWI is robust regarding the choice of initial velocity model. We apply w-AWI tosynthetic and real data obtaining encouraging near-surface imaging results