Bayesian Generalized Nonlinear Models (BGNLM) offer a flexible alternative to GLM while still providing better interpretability than machine learning techniques such as neural networks. In BGNLM, the methods of Bayesian Variable Selection and Model Averaging are applied in an extended GLM setting. Models are fitted to data using MCMC within a genetic framework in an algorithm called GMJMCMC. In this thesis, we present a new implementation of the algorithm as a package in the programming language R. We also present a novel algorithm called S-IRLS-SGD for estimating the MLE of a GLM by subsampling the data. Finally, we present some theory combining the novel algorithm with GMJMCMC/MJMCMC/MCMC and a number of experiments demonstrating the performance of the contributed algorithm.

This paper presents a new approach to the feature-matching problem for 3D reconstruction by taking advantage of GPS and IMU data, along with a prior calibrated stereo camera system. It is expected that pose estimates and calibration can be used to increase feature matching speed and accuracy. Given pose estimates of cameras and extracted features from images, the algorithm first enumerates feature matches based on stereo projection constraints in 2D and then backprojects them to 3D. Then, a grid search algorithm over potential camera poses is proposed to match the 3D features and find the largest group of 3D feature matches between pairs of stereo frames. This approach will provide pose accuracy to within the space that each grid region covers. Further refinement of relative camera poses is performed with an iteratively re-weighted least squares (IRLS) method in order to reject outliers in the 3D matches. The algorithm is shown to be capable of running in real-time correctly, where the majority of processing time is taken by feature extraction and description. The method is shown to outperform standard open source software for reconstruction from imagery. / Master of Science

Nous reportons dans ces travaux de thèse le développement de deux spectromètres à lasers fonctionnant dans la région spectrale du moyen infrarouge (2,78 µm et 8 µm) correspondant aux deux fenêtres atmosphériques pour la détection de traces de gaz. Le premier spectromètre, basé sur la génération de différence de fréquences (DFG) vers 2,78 µm, est couplé à un spectromètre utilisant un laser à cascade quantique (QCL) vers 8 µm dans une cellule multipassages. Ce montage croisé nous a permis de déterminer pour la première fois expérimentalement les intensités de 31 raies d’absorption les plus intenses de la branche Q de la bande fondamentale ν₁ de l’isomère trans de l’acide nitreux (trans-HONO), considéré comme espèce clé pour la capacité d'oxydation atmosphérique. Nous avons exploité le spectromètre à QCL lors d’une campagne de mesures ciblée sur la surveillance continue du méthane (CH₄) pendant le mois de janvier 2013 à Dunkerque. Les observations de la variation de la concentration du CH₄ ont été analysées à l'aide des paramètres météorologiques simultanément enregistrées. Face au besoin d’identification de ses sources d’émission, nous avons développé la technique IRLS (Isotope Ratio Laser Spectrometry) pour la mesure du taux isotopique de ¹³CH₄/¹²CH₄. Les résultats préliminaires sont présentés. / I report in this PhD thesis on the development of two mid-infrared laser spectrometers, based on difference-frequency generation (DFG) and quantum cascade laser (QCL), for application to trace gas monitoring. The DFG spectrometer (2.78 µm) was coupled with the QCL spectrometer (8 µm) to simultaneously measure nitrous acid (HONO) absorption spectra of the v₁ and v₃ bands respectively. Such crossing measurements allow us to determine experimentally, for the first time, the line strengths of 31 absorption lines of the ν1 band of trans isomer of nitrous acid that significantly impacts the air quality and climate change because of its crucial role in the atmospheric oxidation capacity. The QCL spectrometer is also deployed for continuous monitoring of methane (CH₄) during January 2013 in Dunkirk. Methane concentration variation is analyzed with the help of the simultaneously recorded meteorological parameters. In order to identify the sources of CH₄ emission, I developed an Isotope Ratio Laser Spectrometry (IRLS) technique to measure the isotopic ratio of ¹³CH₄/¹²CH₄. Preliminary results are presented.

Introduction of this thesis is dedicated to the description of basic concepts and algorithms for image processing using sparse representation. Furthermore there is mentioned neural network model called Restricted Boltzmann machine, which is in the practical part of the thesis subject of behaving observation in the task of determining the optimal block size for extrapolation using K-SVD algorithm