Real-Time Automatic Object Classification and Tracking using Genetic Programming and NVIDIA R CUDA TM

Genetic Programming (GP) is a widely used methodology for solving various computational problems. GP's problem solving ability is usually hindered by its long execution times. In this thesis, GP is applied toward real-time computer vision. In particular, object classification and tracking using a parallel GP system is discussed. First, a study of suitable GP languages for object classification is presented. Two main GP approaches for visual pattern classification, namely the block-classifiers and the pixel-classifiers, were studied. Results showed that the pixel-classifiers generally performed better. Using these results, a suitable language was selected for the real-time implementation. Synthetic video data was used in the experiments. The goal of the experiments was to evolve a unique classifier for each texture pattern that existed in the video. The experiments revealed that the system was capable of correctly tracking the textures in the video. The performance of the system was on-par with real-time requirements.

Network Similarity Measures and Automatic Construction of Graph Models using Genetic Programming

A complex network is an abstract representation of an intricate system of interrelated elements where the patterns of connection hold significant meaning. One particular complex network is a social network whereby the vertices represent people and edges denote their daily interactions. Understanding social network dynamics can be vital to the mitigation of disease spread as these networks model the interactions, and thus avenues of spread, between individuals. To better understand complex networks, algorithms which generate graphs exhibiting observed properties of real-world networks, known as graph models, are often constructed. While various efforts to aid with the construction of graph models have been proposed using statistical and probabilistic methods, genetic programming (GP) has only recently been considered. However, determining that a graph model of a complex network accurately describes the target network(s) is not a trivial task as the graph models are often stochastic in nature and the notion of similarity is dependent upon the expected behavior of the network. This thesis examines a number of well-known network properties to determine which measures best allowed networks generated by different graph models, and thus the models themselves, to be distinguished. A proposed meta-analysis procedure was used to demonstrate how these network measures interact when used together as classifiers to determine network, and thus model, (dis)similarity. The analytical results form the basis of the fitness evaluation for a GP system used to automatically construct graph models for complex networks. The GP-based automatic inference system was used to reproduce existing, well-known graph models as well as a real-world network. Results indicated that the automatically inferred models exemplified functional similarity when compared to their respective target networks. This approach also showed promise when used to infer a model for a mammalian brain network.

Object-Oriented Genetic Programming for the Automatic Inference of Graph Models for Complex Networks

Complex networks are systems of entities that are interconnected through meaningful relationships. The result of the relations between entities forms a structure that has a statistical complexity that is not formed by random chance. In the study of complex networks, many graph models have been proposed to model the behaviours observed. However, constructing graph models manually is tedious and problematic. Many of the models proposed in the literature have been cited as having inaccuracies with respect to the complex networks they represent. However, recently, an approach that automates the inference of graph models was proposed by Bailey [10] The proposed methodology employs genetic programming (GP) to produce graph models that approximate various properties of an exemplary graph of a targeted complex network. However, there is a great deal already known about complex networks, in general, and often specific knowledge is held about the network being modelled. The knowledge, albeit incomplete, is important in constructing a graph model. However it is difficult to incorporate such knowledge using existing GP techniques. Thus, this thesis proposes a novel GP system which can incorporate incomplete expert knowledge that assists in the evolution of a graph model. Inspired by existing graph models, an abstract graph model was developed to serve as an embryo for inferring graph models of some complex networks. The GP system and abstract model were used to reproduce well-known graph models. The results indicated that the system was able to evolve models that produced networks that had structural similarities to the networks generated by the respective target models.

Inverse Illumination Design with Genetic Programming

Interior illumination is a complex problem involving numerous interacting factors. This research applies genetic programming towards problems in illumination design. The Radiance system is used for performing accurate illumination simulations. Radiance accounts for a number of important environmental factors, which we exploit during fitness evaluation. Illumination requirements include local illumination intensity from natural and artificial sources, colour, and uniformity. Evolved solutions incorporate design elements such as artificial lights, room materials, windows, and glass properties. A number of case studies are examined, including many-objective problems involving up to 7 illumination requirements, the design of a decorative wall of lights, and the creation of a stained-glass window for a large public space. Our results show the technical and creative possibilities of applying genetic programming to illumination design.

Cyclotetrasiloxane Frameworks for the Chemoenzymatic Synthesis of Oligoesters

Immobilized lipase B from Candida antarctica (Novozym® 435, N435) was utilized as part of a chemoenzymatic strategy for the synthesis of branched polyesters based on a cyclotetrasiloxane core in the absence of solvent. Nuclear magnetic resonance spectroscopy and matrix-assisted laser desorption ionization time-of-flight mass spectrometry were utilized to monitor the reactions between tetraester cyclotetrasiloxanes and aliphatic diols. The enzyme-mediated esterification reactions can achieve 65– 80% consumption of starting materials in 24–48 h. Longer reaction times, 72–96 h, resulted in the formation of cross-linked gel-like networks. Gel permeation chromatography of the polymers indicated that the masses were Mw ¼ 11 400, 13 100, and 19 400 g mol 1 for the substrate pairs of C7D4 ester/ octane-1,8-diol, C10D4 ester/pentane-1,5-diol and C10D4 ester/octane-1,8-diol respectively, after 48 h. Extending the polymerization for an additional 24 h with the C10D4 ester/octane-1,8-diol pair gave Mw ¼ 86 800 g mol 1. To the best of our knowledge this represents the first report using lipase catalysis to produce branched polymers that are built from a cyclotetrasiloxane core.

Feature Selection and Classification Using Age Layered Population Structure Genetic Programming

The curse of dimensionality is a major problem in the fields of machine learning, data mining and knowledge discovery. Exhaustive search for the most optimal subset of relevant features from a high dimensional dataset is NP hard. Sub–optimal population based stochastic algorithms such as GP and GA are good choices for searching through large search spaces, and are usually more feasible than exhaustive and deterministic search algorithms. On the other hand, population based stochastic algorithms often suffer from premature convergence on mediocre sub–optimal solutions. The Age Layered Population Structure (ALPS) is a novel metaheuristic for overcoming the problem of premature convergence in evolutionary algorithms, and for improving search in the fitness landscape. The ALPS paradigm uses an age–measure to control breeding and competition between individuals in the population. This thesis uses a modification of the ALPS GP strategy called Feature Selection ALPS (FSALPS) for feature subset selection and classification of varied supervised learning tasks. FSALPS uses a novel frequency count system to rank features in the GP population based on evolved feature frequencies. The ranked features are translated into probabilities, which are used to control evolutionary processes such as terminal–symbol selection for the construction of GP trees/sub-trees. The FSALPS metaheuristic continuously refines the feature subset selection process whiles simultaneously evolving efficient classifiers through a non–converging evolutionary process that favors selection of features with high discrimination of class labels. We investigated and compared the performance of canonical GP, ALPS and FSALPS on high–dimensional benchmark classification datasets, including a hyperspectral image. Using Tukey’s HSD ANOVA test at a 95% confidence interval, ALPS and FSALPS dominated canonical GP in evolving smaller but efficient trees with less bloat expressions. FSALPS significantly outperformed canonical GP and ALPS and some reported feature selection strategies in related literature on dimensionality reduction.

Feature Selection and Classification Using Age Layered Population Structure Genetic Programming

The curse of dimensionality is a major problem in the fields of machine learning, data mining and knowledge discovery. Exhaustive search for the most optimal subset of relevant features from a high dimensional dataset is NP hard. Sub–optimal population based stochastic algorithms such as GP and GA are good choices for searching through large search spaces, and are usually more feasible than exhaustive and determinis- tic search algorithms. On the other hand, population based stochastic algorithms often suffer from premature convergence on mediocre sub–optimal solutions. The Age Layered Population Structure (ALPS) is a novel meta–heuristic for overcoming the problem of premature convergence in evolutionary algorithms, and for improving search in the fitness landscape. The ALPS paradigm uses an age–measure to control breeding and competition between individuals in the population. This thesis uses a modification of the ALPS GP strategy called Feature Selection ALPS (FSALPS) for feature subset selection and classification of varied supervised learning tasks. FSALPS uses a novel frequency count system to rank features in the GP population based on evolved feature frequencies. The ranked features are translated into probabilities, which are used to control evolutionary processes such as terminal–symbol selection for the construction of GP trees/sub-trees. The FSALPS meta–heuristic continuously refines the feature subset selection process whiles simultaneously evolving efficient classifiers through a non–converging evolutionary process that favors selection of features with high discrimination of class labels. We investigated and compared the performance of canonical GP, ALPS and FSALPS on high–dimensional benchmark classification datasets, including a hyperspectral image. Using Tukey’s HSD ANOVA test at a 95% confidence interval, ALPS and FSALPS dominated canonical GP in evolving smaller but efficient trees with less bloat expressions. FSALPS significantly outperformed canonical GP and ALPS and some reported feature selection strategies in related literature on dimensionality reduction.

Optimizing Computational Frameworks to Study the Influence of the Protein Environment on the Individual Site Energies of Chromophores in Photosystem II of Photosynthesis

Photosynthesis is a process in which electromagnetic radiation is converted into chemical energy. Photosystems capture photons with chromophores and transfer their energy to reaction centers using chromophores as a medium. In the reaction center, the excitation energy is used to perform chemical reactions. Knowledge of chromophore site energies is crucial to the understanding of excitation energy transfer pathways in photosystems and the ability to compute the site energies in a fast and accurate manner is mandatory for investigating how protein dynamics ef-fect the site energies and ultimately energy pathways with time. In this work we developed two software frameworks designed to optimize the calculations of chro-mophore site energies within a protein environment. The first is for performing quantum mechanical energy optimizations on molecules and the second is for com-puting site energies of chromophores in a fast and accurate manner using the polar-izability embedding method. The two frameworks allow for the fast and accurate calculation of chromophore site energies within proteins, ultimately allowing for the effect of protein dynamics on energy pathways to be studied. We use these frame-works to compute the site energies of the eight chromophores in the reaction center of photosystem II (PSII) using a 1.9 Å resolution x-ray structure of photosystem II. We compare our results to conflicting experimental data obtained from both isolat-ed intact PSII core preparations and the minimal reaction center preparation of PSII, and find our work more supportive of the former.

A Compiler for the dependently typed language Beluga

Les structures avec des lieurs sont très communes en informatique. Les langages de programmation et les systèmes logiques sont des exemples de structures avec des lieurs. La manipulation de lieurs est délicate, de sorte que l’écriture de programmes qui ma- nipulent ces structures tirerait profit d’un soutien spécifique pour les lieurs. L’environ- nement de programmation Beluga est un exemple d’un tel système. Nous développons et présentons ici un compilateur pour ce système. Parmi les programmes pour lesquels Beluga est spécialement bien adapté, plusieurs peuvent bénéficier d’un compilateur. Par exemple, les programmes pour valider les types (les "type-checkers"), les compilateurs et les interpréteurs tirent profit du soutien spécifique des lieurs et des types dépendants présents dans le langage. Ils nécessitent tous également une exécution efficace, que l’on propose d’obtenir par le biais d’un compilateur. Le but de ce travail est de présenter un nouveau compilateur pour Beluga, qui emploie une représentation interne polyvalente et permet de partager du code entre plusieurs back-ends. Une contribution notable est la compilation du filtrage de Beluga, qui est particulièrement puissante dans ce langage.

Approches générales de résolution pour les problèmes multi-attributs de tournées de véhicules et confection d'horaires

Thèse réalisée en cotutelle entre l'Université de Montréal et l'Université de Technologie de Troyes

Evaluation of mitochondrial function in a model of developmental programming of hypertension associated with transient neonatal oxygen exposure

UNE EXPOSITION NÉONATALE À L’OXYGÈNE MÈNE À DES MODIFICATIONS DE LA FONCTION MITOCHONDRIALE CHEZ LE RAT ADULTE Introduction: L’exposition à l’oxygène (O2) des ratons nouveau-nés a des conséquences à l’âge adulte dont une hypertension artérielle (HTA), une dysfonction vasculaire, une néphropénie et des indices de stress oxydant. En considérant que les reins sont encore en développement actif lors des premiers jours après la naissance chez les rats, jouent un rôle clé dans le développement de l’hypertension et qu’une dysfonction mitochondriale est associé à une augmentation du stress oxydant, nous postulons que les conditions délétères néonatales peuvent avoir un impact significatif au niveau rénal sur la modulation de l’expression de protéines clés du fonctionnement mitochondrial et une production mitochondriale excessive d’espèces réactives de l’ O2. Méthodes: Des ratons Sprague-Dawley sont exposés à 80% d’O2 (H) ou 21% O2 (Ctrl) du 3e au 10e jr de vie. En considérant que plusieurs organes des rats sont encore en développement actif à la naissance, ces rongeurs sont un modèle reconnu pour étudier les complications d’une hyperoxie néonatale, comme celles liées à une naissance prématurée chez l’homme. À 4 et à 16 semaines, les reins sont prélevés et les mitochondries sont extraites suivant une méthode d’extraction standard, avec un tampon contenant du sucrose 0.32 M et différentes centrifugations. L’expression des protéines mitochondriales a été mesurée par Western blot, tandis que la production d’ H202 et les activités des enzymes clés du cycle de Krebs ont été évaluées par spectrophotométrie. Les résultats sont exprimés par la moyenne ± SD. Résultats: Les rats mâles H de 16 semaines (n=6) présentent une activité de citrate synthase (considéré standard interne de l’expression protéique et de l’abondance mitochondriales) augmentée (12.4 ± 8.4 vs 4.1 ± 0.5 μmole/mL/min), une diminution de l’activité d’aconitase (enzyme sensible au redox mitochondrial) (0.11 ± 0.05 vs 0.20 ± 0.04 μmoles/min/mg mitochondrie), ainsi qu’une augmentation dans la production de H202 (7.0 ± 1.3 vs 5.4 ± 0.8 ρmoles/mg protéines mitochondriales) comparativement au groupe Ctrl (n=6 mâles et 4 femelles). Le groupe H (vs Ctrl) présente également une diminution dans l’expression de peroxiredoxin-3 (Prx3) (H 0.61±0.06 vs. Ctrl 0.78±0.02 unité relative, -23%; p<0.05), une protéine impliquée dans l’élimination d’ H202, de l’expression du cytochrome C oxidase (Complexe IV) (H 1.02±0.04 vs. Ctrl 1.20±0.02 unité relative, -15%; p<0.05), une protéine de la chaine de respiration mitochondriale, tandis que l’expression de la protéine de découplage (uncoupling protein)-2 (UCP2), impliquée dans la dispersion du gradient proton, est significativement augmentée (H 1.05±0.02 vs. Ctrl 0.90±0.03 unité relative, +17%; p<0.05). Les femelles H (n=6) (vs Ctrl, n=6) de 16 semaines démontrent une augmentation significative de l’activité de l’aconitase (0.33±0.03 vs 0.17±0.02 μmoles/min/mg mitochondrie), de l’expression de l’ATP synthase sous unité β (H 0.73±0.02 vs. Ctrl 0.59±0.02 unité relative, +25%; p<0.05) et de l’expression de MnSOD (H 0.89±0.02 vs. Ctrl 0.74±0.03 unité relative, +20%; p<0.05) (superoxide dismutase mitochondriale, important antioxidant), tandis que l’expression de Prx3 est significativement réduite (H 1.1±0.07 vs. Ctrl 0.85±0.01 unité relative, -24%; p<0.05). À 4 semaines, les mâles H (vs Ctrl) présentent une augmentation significative de l’expression de Prx3 (H 0.72±0.03 vs. Ctrl 0.56±0.04 unité relative, +31%; p<0.05) et les femelles présentent une augmentation significative de l’expression d’UCP2 (H 1.22±0.05 vs. Ctrl 1.03±0.04 unité relative, +18%; p<0.05) et de l’expression de MnSOD (H 1.36±0.01 vs. 1.19±0.06 unité relative, +14%; p<0.05). Conclusions: Une exposition néonatale à l’O2 chez le rat adulte mène à des indices de dysfonction mitochondriale dans les reins adultes, associée à une augmentation dans la production d’espèces réactives de l’oxygène, suggérant que ces modifications mitochondriales pourraient jouer un rôle dans l’hypertension artérielle et d’un stress oxydant, et par conséquent, être un facteur possible dans la progression vers des maladies cardiovasculaires. Mots-clés: Mitochondries, Reins, Hypertension, Oxygène, Stress Oxydant, Programmation

Lagrangian-informed mixed integer programming reformulations

La programmation linéaire en nombres entiers est une approche robuste qui permet de résoudre rapidement de grandes instances de problèmes d'optimisation discrète. Toutefois, les problèmes gagnent constamment en complexité et imposent parfois de fortes limites sur le temps de calcul. Il devient alors nécessaire de développer des méthodes spécialisées afin de résoudre approximativement ces problèmes, tout en calculant des bornes sur leurs valeurs optimales afin de prouver la qualité des solutions obtenues. Nous proposons d'explorer une approche de reformulation en nombres entiers guidée par la relaxation lagrangienne. Après l'identification d'une forte relaxation lagrangienne, un processus systématique permet d'obtenir une seconde formulation en nombres entiers. Cette reformulation, plus compacte que celle de Dantzig et Wolfe, comporte exactement les mêmes solutions entières que la formulation initiale, mais en améliore la borne linéaire: elle devient égale à la borne lagrangienne. L'approche de reformulation permet d'unifier et de généraliser des formulations et des méthodes de borne connues. De plus, elle offre une manière simple d'obtenir des reformulations de moins grandes tailles en contrepartie de bornes plus faibles. Ces reformulations demeurent de grandes tailles. C'est pourquoi nous décrivons aussi des méthodes spécialisées pour en résoudre les relaxations linéaires. Finalement, nous appliquons l'approche de reformulation à deux problèmes de localisation. Cela nous mène à de nouvelles formulations pour ces problèmes; certaines sont de très grandes tailles, mais nos méthodes de résolution spécialisées les rendent pratiques.

Modeling High-Dimensional Audio Sequences with Recurrent Neural Networks

Cette thèse étudie des modèles de séquences de haute dimension basés sur des réseaux de neurones récurrents (RNN) et leur application à la musique et à la parole. Bien qu'en principe les RNN puissent représenter les dépendances à long terme et la dynamique temporelle complexe propres aux séquences d'intérêt comme la vidéo, l'audio et la langue naturelle, ceux-ci n'ont pas été utilisés à leur plein potentiel depuis leur introduction par Rumelhart et al. (1986a) en raison de la difficulté de les entraîner efficacement par descente de gradient. Récemment, l'application fructueuse de l'optimisation Hessian-free et d'autres techniques d'entraînement avancées ont entraîné la recrudescence de leur utilisation dans plusieurs systèmes de l'état de l'art. Le travail de cette thèse prend part à ce développement. L'idée centrale consiste à exploiter la flexibilité des RNN pour apprendre une description probabiliste de séquences de symboles, c'est-à-dire une information de haut niveau associée aux signaux observés, qui en retour pourra servir d'à priori pour améliorer la précision de la recherche d'information. Par exemple, en modélisant l'évolution de groupes de notes dans la musique polyphonique, d'accords dans une progression harmonique, de phonèmes dans un énoncé oral ou encore de sources individuelles dans un mélange audio, nous pouvons améliorer significativement les méthodes de transcription polyphonique, de reconnaissance d'accords, de reconnaissance de la parole et de séparation de sources audio respectivement. L'application pratique de nos modèles à ces tâches est détaillée dans les quatre derniers articles présentés dans cette thèse. Dans le premier article, nous remplaçons la couche de sortie d'un RNN par des machines de Boltzmann restreintes conditionnelles pour décrire des distributions de sortie multimodales beaucoup plus riches. Dans le deuxième article, nous évaluons et proposons des méthodes avancées pour entraîner les RNN. Dans les quatre derniers articles, nous examinons différentes façons de combiner nos modèles symboliques à des réseaux profonds et à la factorisation matricielle non-négative, notamment par des produits d'experts, des architectures entrée/sortie et des cadres génératifs généralisant les modèles de Markov cachés. Nous proposons et analysons également des méthodes d'inférence efficaces pour ces modèles, telles la recherche vorace chronologique, la recherche en faisceau à haute dimension, la recherche en faisceau élagué et la descente de gradient. Finalement, nous abordons les questions de l'étiquette biaisée, du maître imposant, du lissage temporel, de la régularisation et du pré-entraînement.

Self-assembled Transition Metal Coordination Frameworks of Carbohydrazone and Thiocarbohydrazone Ligands: Structural, Spectral, Magnetic and Anticancer Properties

Chemistry occupies a unique middle position in the scientific arena, between physics and mathematics on the one side and biology, ecology, sociology and economics on the other [1]. Chemistry is the science of matter and of its transformations, and life is its highest expression [2]. According to reductionist thinking biology is reducible into chemistry, chemistry into physics, and ultimately physics into mathematics. Reductionism implies the ease of understanding one level in terms of another.The work presented this thesis comprises synthesis and characterization of suitably substituted thiocarbohydrazone and carbohydrazone ligand building blocks, self-assembled metallosupramolecular square grid complexes as well as some di/multinuclear complexes. The primary aim was the deliberate syntheses of some novel transition metal framework complexes, mainly metallosupramolecular coordination square grids by self-assembly and their physico-chemical characterization. The work presented, however, also include synthesis and characterization of four mononuclear Ni(II) complexes of two thiosemicarbazones, which we carried out as a preliminary and supporting study. Based on the present work we would like to conclude that the carbohydrazones, thiocarbohydrazones and their coordination framework complexes of transition metals are promising systems for wide application in science and technology varied from physics to biotechnology. Novel classes of materials and biologically important potential compounds open up further scope of researches and we hopefully welcome any sort of related research to make this work more valuable.