Much has been said and written over recent years about Attention-Deficit/Hyperactivity Disorder. There is a certain amount of confusion as to what exactly the condition constitutes and controversy continues to rage regarding treatment. A significant number of children appear to be affected. Previously, parents and teachers ·were blamed for failing to discipline effectively. Often, the difficulties remained undiagnosed and untreated. Thanks to the wisdom of so many experts who have generously shared their knowledge and considerable expertise, there is an increased awareness of ADHD. Although there is no cure, there are ways to manage the difficulties. However, early diagnosis and intervention is critical. Since many different symptoms are associated with the disorder, a multi-modal treatment plan has been found to lead to a better outcome. For the purpose of this study, a multi-modal programme was planned to address the needs of a small group of children with ADHD and their parents. The intention was to empower the parents, within a supportive group environment, by providing them with knowledge about the disorder and guidelines for managing the difficult behaviour. In addition, an attempt was made to change the negative behaviour patterns of the children through the medium of story-telling. It was hoped that by reducing the levels of parental stress, parents would be more competent to cope with their educational demands, so that their children could be guided more positively towards adulthood. The results of the programme were positive. Teachers and parents reported better behaviour by the children. The parents' stress levels were reduced. The parents expressed greater understanding about the disorder and a hopefulness that they could better manage their children. They felt they had benefitted from the advice given by other parents who were facing similar challenges. However, they felt that a short-term programme was insufficient to address all their needs and they expressed a need for ongoing support. In view of the chronicity of the disorder and the constantly changing needs of the child on his journey towards adulthood, cognisance was taken of the fact that longterm intervention is essential. / Psychology of Education / D.Ed. (Psychology of Education)

Mirrors and Vanities is a multi-modal collection which showcases the diversity of working in long and short storytelling forms. Featured in this thesis are fiction, nonfiction, graphic narrative, and screenplay. Using unconventional approaches to storytelling in order to achieve emotional resonance with the audience while maintaining high standards for craft, these stories and essays explore the costs inherent to the subtle nuances of interpersonal relationships. The fiction focuses on the complications of characters keeping secrets. A husband discovers the truth behind his wife’s miscarriage. A girl visits her fiancé in purgatory. A boy crosses a line and loses his best friend. Meanwhile, the nonfiction centers on self-discovery and gender roles associated with power struggles. A schizophrenic threatens to ruin my mother’s wedding. I rediscover my relationship with my father through food writing. Sword-work teaches me to fail and succeed at making martial art. The title work of the thesis is a collaged story highlighting the tribulations of a physicist fixated on recovering his lost love by manipulating the multiverse. The multi-modal format implicates the nebulosity of physics theories and how different aspects of the narrative can be presented in various formats to best suit the nature of the storytelling. Through the interactions of characters in mundane and extraordinary circumstances, the works in this thesis examine the consequences of choice, the contrast between reality and expectation, coming of age, and the Truth of narrative.

Ces dernières années, le domaine de l’apprentissage profond a connu des progrès énormes dans des applications allant de la génération d’images, détection d’objets, modélisation du langage à la réponse aux questions visuelles. Les approches classiques telles que l’apprentissage supervisé nécessitent de grandes quantités de données étiquetées et spécifiques à la tâches. Cependant, celles-ci sont parfois coûteuses, peu pratiques, ou trop longues à collecter. La modélisation efficace en données, qui comprend des techniques comme l’apprentissage few-shot (à partir de peu d’exemples) et l’apprentissage self-supervised (auto-supervisé), tentent de remédier au manque de données spécifiques à la tâche en exploitant de grandes quantités de données plus “générales”. Les progrès de l’apprentissage profond, et en particulier de l’apprentissage few-shot, s’appuient sur les benchmarks (suites d’évaluation), les métriques d’évaluation et les jeux de données, car ceux-ci sont utilisés pour tester et départager différentes méthodes sur des tâches précises, et identifier l’état de l’art. Cependant, du fait qu’il s’agit de versions idéalisées de la tâche à résoudre, les benchmarks sont rarement équivalents à la tâche originelle, et peuvent avoir plusieurs limitations qui entravent leur rôle de sélection des directions de recherche les plus prometteuses. De plus, la définition de métriques d’évaluation pertinentes peut être difficile, en particulier dans le cas de sorties structurées et en haute dimension, telles que des images, de l’audio, de la parole ou encore du texte. Cette thèse discute des limites et des perspectives des benchmarks existants, des fonctions de coût (training losses) et des métriques d’évaluation (evaluation metrics), en mettant l’accent sur la modélisation générative - les Réseaux Antagonistes Génératifs (GANs) en particulier - et la modélisation efficace des données, qui comprend l’apprentissage few-shot et self-supervised. La première contribution est une discussion de la tâche de modélisation générative, suivie d’une exploration des propriétés théoriques et empiriques des fonctions de coût des GANs. La deuxième contribution est une discussion sur la limitation des few-shot classification benchmarks, certains ne nécessitant pas de généralisation à de nouvelles sémantiques de classe pour être résolus, et la proposition d’une méthode de base pour les résoudre sans étiquettes en phase de testing. La troisième contribution est une revue sur les méthodes few-shot et self-supervised de détection d’objets , qui souligne les limites et directions de recherche prometteuses. Enfin, la quatrième contribution est une méthode efficace en données pour la description de vidéo qui exploite des jeux de données texte et vidéo non supervisés. / In recent years, the field of deep learning has seen tremendous progress for applications ranging from image generation, object detection, language modeling, to visual question answering. Classic approaches such as supervised learning require large amounts of task-specific and labeled data, which may be too expensive, time-consuming, or impractical to collect. Data-efficient methods, such as few-shot and self-supervised learning, attempt to deal with the limited availability of task-specific data by leveraging large amounts of general data. Progress in deep learning, and in particular, few-shot learning, is largely driven by the relevant benchmarks, evaluation metrics, and datasets. They are used to test and compare different methods on a given task, and determine the state-of-the-art. However, due to being idealized versions of the task to solve, benchmarks are rarely equivalent to the original task, and can have several limitations which hinder their role of identifying the most promising research directions. Moreover, defining meaningful evaluation metrics can be challenging, especially in the case of high-dimensional and structured outputs, such as images, audio, speech, or text. This thesis discusses the limitations and perspectives of existing benchmarks, training losses, and evaluation metrics, with a focus on generative modeling—Generative Adversarial Networks (GANs) in particular—and data-efficient modeling, which includes few-shot and self-supervised learning. The first contribution is a discussion of the generative modeling task, followed by an exploration of theoretical and empirical properties of the GAN loss. The second contribution is a discussion of a limitation of few-shot classification benchmarks, which is that they may not require class semantic generalization to be solved, and the proposal of a baseline method for solving them without test-time labels. The third contribution is a survey of few-shot and self-supervised object detection, which points out the limitations and promising future research for the field. Finally, the fourth contribution is a data-efficient method for video captioning, which leverages unsupervised text and video datasets, and explores several multimodal pretraining strategies.

Membrane filtration processes are widely utilized across different industrial sectors for biological and environmental separations. Examples of the former are sterile filtration and protein fractionation via microfiltration (MF) and ultrafiltration (UF) while drinking water treatment, tertiary treatment of wastewater, water reuse and desalination via MF, UF, nanofiltration (NF) and reverse-osmosis (RO) are examples of the latter. A common misconception is that the performance of membrane separation is solely dependent on the membrane pore size, whereas a multitude of parameters including solution conditions, solute concentration, presence of specific ions, hydrodynamic conditions, membrane structure and surface properties can significantly influence the separation performance and the membrane’s fouling propensity. The conventional approach for studying filtration performance is to use a single lab- or pilot-scale module and perform numerous experiments in a sequential manner which is both time-consuming and requires large amounts of material. Alternatively, high-throughput (HT) techniques, defined as the miniaturized version of conventional unit operations which allow for multiple experiments to be run in parallel and require a small amount of sample, can be employed. There is a growing interest in the use of HT techniques to speed up the testing and optimization of membrane-based separations. In this work, different HT screening approaches are developed and utilized for the evaluation and optimization of filtration performance using flat-sheet and hollow-fiber (HF) membranes used in biological and environmental separations. The effects of various process factors were evaluated on the separation of different biomolecules by combining a HT filtration method using flat-sheet UF membranes and design-of-experiments methods. Additionally, a novel HT platform was introduced for multi-modal (constant transmembrane pressure vs. constant flux) testing of flat-sheet membranes used in bio-separations. Furthermore, the first-ever HT modules for parallel testing of HF membranes were developed for rapid fouling tests as well as extended filtration evaluation experiments. The usefulness of the modules was demonstrated by evaluating the filtration performance of different foulants under various operating conditions as well as running surface modification experiments. The techniques described herein can be employed for rapid determination of the optimal combination of conditions that result in the best filtration performance for different membrane separation applications and thus eliminate the need to perform numerous conventional lab-scale tests. Overall, more than 250 filtration tests and 350 hydraulic permeability measurements were performed and analyzed using the HT platforms developed in this thesis. / Thesis / Doctor of Philosophy (PhD) / Membrane filtration is widely used as a key separation process in different industries. For example, microfiltration (MF) and ultrafiltration (UF) are used for sterilization and purification of bio-products. Furthermore, MF, UF and reverse-osmosis (RO) are used for drinking water and wastewater treatment. A common misconception is that membrane filtration is a process solely based on the pore size of the membrane whereas numerous factors can significantly affect the performance. Conventionally, a large number of lab- or full-scale experiments are performed to find the optimum operating conditions for each filtration process. High-throughput (HT) techniques are powerful methods to accelerate the pace of process optimization—they allow for multiple experiments to be run in parallel and require smaller amounts of sample. This thesis focuses on the development of different HT techniques that require a minimal amount of sample for parallel testing and optimization of membrane filtration processes with applications in environmental and biological separations. The introduced techniques can reduce the amount of sample used in each test between 10-50 times and accelerate process development and optimization by running parallel tests.